Initial revision

195 files changed, 4970 insertions, 6074 deletions

diff --git a/arch/alpha/defconfig b/arch/alpha/defconfig
index a6ee5a17c..0c3c65d28 100644
--- a/arch/alpha/defconfig
+++ b/arch/alpha/defconfig
@@ -42,6 +42,13 @@ CONFIG_SYSVIPC=y
 CONFIG_SYSCTL=y
 CONFIG_BINFMT_AOUT=y
 CONFIG_BINFMT_ELF=y
+CONFIG_BINFMT_EM86=y
+# CONFIG_PNP_PARPORT is not set
+
+#
+# Plug and Play support
+#
+# CONFIG_PNP is not set
 
 #
 # Floppy, IDE, and other block devices
@@ -72,9 +79,10 @@ CONFIG_BLK_DEV_RAM=y
 # CONFIG_FIREWALL is not set
 # CONFIG_NET_ALIAS is not set
 CONFIG_INET=y
-# CONFIG_IP_FORWARD is not set
 # CONFIG_IP_MULTICAST is not set
 # CONFIG_IP_ACCT is not set
+# CONFIG_IP_ROUTER is not set
+# CONFIG_NET_IPIP is not set
 
 #
 # (it is safe to leave these untouched)
@@ -133,11 +141,11 @@ CONFIG_SCSI_CONSTANTS=y
 # CONFIG_SCSI_NCR53C406A is not set
 # CONFIG_SCSI_NCR53C7xx is not set
 # CONFIG_SCSI_NCR53C8XX is not set
-# CONFIG_SCSI_PPA is not set
 # CONFIG_SCSI_PAS16 is not set
 # CONFIG_SCSI_QLOGIC_FAS is not set
 CONFIG_SCSI_QLOGIC_ISP=y
 # CONFIG_SCSI_SEAGATE is not set
+# CONFIG_SCSI_DC390T is not set
 # CONFIG_SCSI_T128 is not set
 # CONFIG_SCSI_U14_34F is not set
 # CONFIG_SCSI_ULTRASTOR is not set
@@ -155,19 +163,22 @@ CONFIG_NET_ETHERNET=y
 # CONFIG_NET_VENDOR_SMC is not set
 # CONFIG_NET_ISA is not set
 CONFIG_NET_EISA=y
+# CONFIG_PCNET32 is not set
 # CONFIG_APRICOT is not set
+# CONFIG_CS89x0 is not set
 CONFIG_DE4X5=y
 # CONFIG_DEC_ELCP is not set
 # CONFIG_DGRS is not set
+# CONFIG_EEXPRESS_PRO100 is not set
 # CONFIG_NET_POCKET is not set
 # CONFIG_FDDI is not set
 # CONFIG_DLCI is not set
-# CONFIG_PLIP is not set
 # CONFIG_PPP is not set
 # CONFIG_NET_RADIO is not set
-# CONFIG_LAPBETHER is not set
 # CONFIG_SLIP is not set
 # CONFIG_TR is not set
+# CONFIG_LAPBETHER is not set
+# CONFIG_X25_ASY is not set
 
 #
 # ISDN subsystem
@@ -192,12 +203,16 @@ CONFIG_MSDOS_FS=y
 CONFIG_PROC_FS=y
 CONFIG_NFS_FS=y
 # CONFIG_ROOT_NFS is not set
+# CONFIG_NFSD is not set
+CONFIG_SUNRPC=y
+CONFIG_LOCKD=y
 # CONFIG_SMB_FS is not set
 CONFIG_ISO9660_FS=y
 # CONFIG_HPFS_FS is not set
 # CONFIG_SYSV_FS is not set
 # CONFIG_AFFS_FS is not set
 # CONFIG_ROMFS_FS is not set
+# CONFIG_AUTOFS_FS is not set
 # CONFIG_UFS_FS is not set
 
 #
@@ -206,12 +221,8 @@ CONFIG_ISO9660_FS=y
 CONFIG_VT=y
 CONFIG_VT_CONSOLE=y
 CONFIG_SERIAL=y
-# CONFIG_DIGI is not set
-# CONFIG_CYCLADES is not set
-# CONFIG_STALDRV is not set
-# CONFIG_RISCOM8 is not set
-# CONFIG_ESPSERIAL is not set
-# CONFIG_PRINTER is not set
+# CONFIG_SERIAL_EXTENDED is not set
+# CONFIG_SERIAL_NONSTANDARD is not set
 CONFIG_MOUSE=y
 # CONFIG_ATIXL_BUSMOUSE is not set
 # CONFIG_BUSMOUSE is not set
@@ -224,7 +235,6 @@ CONFIG_PSMOUSE=y
 # CONFIG_APM is not set
 # CONFIG_WATCHDOG is not set
 # CONFIG_RTC is not set
-CONFIG_RTC_ARC=y
 
 #
 # Sound
diff --git a/arch/alpha/kernel/entry.S b/arch/alpha/kernel/entry.S
index ef582b80e..af26f8996 100644
--- a/arch/alpha/kernel/entry.S
+++ b/arch/alpha/kernel/entry.S
@@ -126,7 +126,8 @@
 .ent	entInt
 entInt:
 	SAVE_ALL
-	ldq	$8,current_set
+	lda	$8,0x3fff
+	bic	$30,$8,$8
 	jsr	$26,do_entInt
 	br	$31,ret_from_sys_call
 .end entInt
@@ -147,7 +148,8 @@ entMM:
 	stq	$15,48($30)
 	addq	$30,56,$19
 /* handle the fault */
-	ldq	$8,current_set
+	lda	$8,0x3fff
+	bic	$30,$8,$8
 	jsr	$26,do_page_fault
 /* reload the registers after the exception code played.  */
 	ldq	$9,0($30)
@@ -167,7 +169,8 @@ entMM:
 .ent	entArith
 entArith:
 	SAVE_ALL
-	ldq	$8,current_set
+	lda	$8,0x3fff
+	bic	$30,$8,$8
 	/* How much of a win is this clockwise?  We are, after all, messing
 	   up the call/return prefetch stack.  -- rth */
 	lda	$27,do_entArith
@@ -180,7 +183,8 @@ entArith:
 .ent	entIF
 entIF:
 	SAVE_ALL
-	ldq	$8,current_set
+	lda	$8,0x3fff
+	bic	$30,$8,$8
 	lda	$27,do_entIF
 	lda	$26,ret_from_sys_call
 	jsr	$31,($27),do_entIF
@@ -221,12 +225,13 @@ kernel_clone:
 .globl	__kernel_thread
 .ent	__kernel_thread
 __kernel_thread:
+	ldgp $29,0($27)		/* we can be called from a module */
 	.frame $30, 4*8, $26
 	subq $30,4*8,$30
 	stq $10,16($30)
 	stq $9,8($30)
 	stq $26,0($30)
-	.prologue 0
+	.prologue 1
 	bis $17,$17,$9		/* save fn */
 	bis $18,$18,$10		/* save arg */
 	bsr $26,kernel_clone
@@ -238,10 +243,9 @@ __kernel_thread:
 	ret $31,($26),1
 /* this is in child: look out as we don't have any stack here.. */
 1:	bis $9,$9,$27		/* get fn */
-	br $29,2f
-2:	ldgp $29,0($29)
+	lda $8,0x3fff
 	bis $10,$10,$16		/* get arg */
-	ldq $8,current_set
+	bic $30,$8,$8		/* get current */
 	jsr $26,($27)
 	bis $0,$0,$16
 	jsr $26,sys_exit
@@ -382,7 +386,8 @@ entUna:
 	stq $29,232($30)
 	stq $30,240($30)
 	stq $31,248($30)
-	ldq $8,current_set
+	lda $8,0x3fff
+	bic $30,$8,$8
 	jsr $26,do_entUna
 	ldq $0,0($30)
 	ldq $1,8($30)
@@ -432,7 +437,8 @@ entUnaUser:
 	stq $14,40($30)
 	stq $15,48($30)
 	bis $31,$30,$19
-	ldq $8,current_set
+	lda $8,0x3fff
+	bic $30,$8,$8
 	jsr $26,do_entUnaUser
 	ldq $9,0($30)
 	ldq $10,8($30)
@@ -497,7 +503,8 @@ alpha_switch_to:
 .ent	entSys
 entSys:
 	SAVE_ALL
-	ldq	$8,current_set
+	lda	$8,0x3fff
+	bic	$30,$8,$8
 	lda	$4,NR_SYSCALLS($31)
 	stq	$16,SP_OFF+24($30)
 	lda	$5,sys_call_table
@@ -532,7 +539,7 @@ ret_from_handle_bh:
 ret_from_reschedule:
 	lda	$0,need_resched
 	ldl	$2,0($0)
-	lda	$4,init_task
+	lda	$4,init_task_union
 	bne	$2,reschedule
 	xor	$4,$8,$4
 	beq	$4,restore_all
diff --git a/arch/alpha/kernel/head.S b/arch/alpha/kernel/head.S
index 17ca4581a..f76213624 100644
--- a/arch/alpha/kernel/head.S
+++ b/arch/alpha/kernel/head.S
@@ -23,9 +23,11 @@ _stext:
 __start:
 	br	$27,1f
 1:	ldgp	$29,0($27)
-	/* We need to get current loaded up with our first task.  */
-	ldq	$8,current_set
-	/* And then we can start the kernel.  */
+	/* We need to get current loaded up with our first task...  */
+	lda	$8,init_task_union
+	/* ... and find our stack ... */
+	lda	$30,0x4000($8)
+	/* ... and then we can start the kernel.  */
 	jsr	$26,start_kernel
 	halt
 	.end __start
@@ -63,22 +65,6 @@ rdusp:
 	.end rdusp
 
 	.align 3
-	.globl	tbi
-	.ent	tbi
-tbi:
-	call_pal PAL_tbi
-	ret	($26)
-	.end tbi
-
-	.align 3
-	.globl	imb
-	.ent	imb
-imb:
-	call_pal PAL_imb
-	ret	($26)
-	.end imb
-
-	.align 3
 	.globl	rdmces
 	.ent	rdmces
 rdmces:
diff --git a/arch/alpha/kernel/process.c b/arch/alpha/kernel/process.c
index 2eb2d51b7..b6c97e726 100644
--- a/arch/alpha/kernel/process.c
+++ b/arch/alpha/kernel/process.c
@@ -37,6 +37,24 @@
 #include <asm/uaccess.h>
 #include <asm/system.h>
 #include <asm/io.h>
+#include <asm/pgtable.h>
+
+/*
+ * Initial task structure. Make this a per-architecture thing,
+ * because different architectures tend to have different
+ * alignment requirements and potentially different initial
+ * setup.
+ */
+
+unsigned long init_user_stack[1024] = { STACK_MAGIC, };
+static struct vm_area_struct init_mmap = INIT_MMAP;
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS;
+struct mm_struct init_mm = INIT_MM;
+
+union task_union init_task_union __attribute__((section("init_task")))
+	 = { task: INIT_TASK };
 
 /*
  * No need to acquire the kernel lock, we're entirely local..
@@ -186,7 +204,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 	stack_offset = PAGE_SIZE - sizeof(struct pt_regs);
 	if (!(regs->ps & 8))
 		stack_offset = (PAGE_SIZE-1) & (unsigned long) regs;
-	childregs = (struct pt_regs *) (p->kernel_stack_page + stack_offset);
+	childregs = (struct pt_regs *) (stack_offset + PAGE_SIZE + (unsigned long)p);
 		
 	*childregs = *regs;
 	childregs->r0 = 0;
diff --git a/arch/alpha/kernel/ptrace.c b/arch/alpha/kernel/ptrace.c
index 019fb6b95..c388f0b51 100644
--- a/arch/alpha/kernel/ptrace.c
+++ b/arch/alpha/kernel/ptrace.c
@@ -67,7 +67,7 @@
  *  |				     |			      |
  *  |				     |			      v
  *  +================================+ <-------------------------
- *					task->kernel_stack_page
+ *					task + PAGE_SIZE
  */
 #define PT_REG(reg)	(PAGE_SIZE - sizeof(struct pt_regs)	\
 			 + (long)&((struct pt_regs *)0)->reg)
@@ -107,19 +107,6 @@ static unsigned short regoff[] = {
 
 static long zero;
 
-
-/* change a pid into a task struct. */
-static inline struct task_struct * get_task(int pid)
-{
-	int i;
-
-	for (i = 1; i < NR_TASKS; i++) {
-		if (task[i] != NULL && (task[i]->pid == pid))
-			return task[i];
-	}
-	return NULL;
-}
-
 /*
  * Get contents of register REGNO in task TASK.
  */
@@ -133,7 +120,7 @@ static inline long get_reg(struct task_struct * task, long regno)
 		zero = 0;
 		addr = &zero;
 	} else {
-		addr = (long *) (task->kernel_stack_page + regoff[regno]);
+		addr = (long *) (regoff[regno] + PAGE_SIZE + (long)task);
 	}
 	return *addr;
 }
@@ -150,7 +137,7 @@ static inline int put_reg(struct task_struct *task, long regno, long data)
 	} else if (regno == 31) {
 		addr = &zero;
 	} else {
-		addr = (long *) (task->kernel_stack_page + regoff[regno]);
+		addr = (long *) (regoff[regno] + PAGE_SIZE + (long)task);
 	}
 	*addr = data;
 	return 0;
@@ -174,7 +161,7 @@ static unsigned long get_long(struct task_struct * tsk,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (pgd_none(*pgdir)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -184,7 +171,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -194,7 +181,7 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
@@ -225,7 +212,7 @@ static void put_long(struct task_struct * tsk, struct vm_area_struct * vma,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (!pgd_present(*pgdir)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -235,7 +222,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -245,12 +232,12 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
 	if (!pte_write(*pgtable)) {
-		do_wp_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 /* this is a hack for non-kernel-mapped video buffers and similar */
@@ -507,7 +494,7 @@ asmlinkage long sys_ptrace(long request, long pid, long addr, long data,
 	if (pid == 1)		/* you may not mess with init */
 		goto out;
 	ret = -ESRCH;
-	if (!(child = get_task(pid)))
+	if (!(child = find_task_by_pid(pid)))
 		goto out;
 	if (request == PTRACE_ATTACH) {
 		ret = -EPERM;
diff --git a/arch/alpha/kernel/setup.c b/arch/alpha/kernel/setup.c
index 0fa1129ee..31a1f21fc 100644
--- a/arch/alpha/kernel/setup.c
+++ b/arch/alpha/kernel/setup.c
@@ -21,9 +21,9 @@
 #include <linux/tty.h>
 #include <linux/delay.h>
 #include <linux/config.h>	/* CONFIG_ALPHA_LCA etc */
+#include <linux/ioport.h>
 
 #ifdef CONFIG_RTC
-#include <linux/ioport.h>
 #include <linux/timex.h>
 #endif
 
diff --git a/arch/alpha/mm/fault.c b/arch/alpha/mm/fault.c
index a7b1e6f2c..a8bc34108 100644
--- a/arch/alpha/mm/fault.c
+++ b/arch/alpha/mm/fault.c
@@ -97,7 +97,7 @@ good_area:
 		if (!(vma->vm_flags & VM_WRITE))
 			goto bad_area;
 	}
-	handle_mm_fault(vma, address, cause > 0);
+	handle_mm_fault(tsk, vma, address, cause > 0);
 	up(&mm->mmap_sem);
 	return;
 
diff --git a/arch/alpha/mm/init.c b/arch/alpha/mm/init.c
index 13e8e4ce4..67faa97d4 100644
--- a/arch/alpha/mm/init.c
+++ b/arch/alpha/mm/init.c
@@ -133,7 +133,6 @@ unsigned long paging_init(unsigned long start_mem, unsigned long end_mem)
 	init_task.tss.ptbr = newptbr;
 	init_task.tss.pal_flags = 1;	/* set FEN, clear everything else */
 	init_task.tss.flags = 0;
-	init_task.kernel_stack_page = INIT_STACK;
 	load_PCB(&init_task.tss);
 
 	flush_tlb_all();
@@ -182,7 +181,7 @@ void free_initmem (void)
                 atomic_set(&mem_map[MAP_NR(addr)].count, 1);
                 free_page(addr);
         }
-        printk ("Freeing unused kernel memory: %dk freed\n",
+        printk ("Freeing unused kernel memory: %ldk freed\n",
 		(&__init_end - &__init_begin) >> 10);
 }
 
diff --git a/arch/alpha/vmlinux.lds b/arch/alpha/vmlinux.lds
index 2a5f00989..0fb2276ea 100644
--- a/arch/alpha/vmlinux.lds
+++ b/arch/alpha/vmlinux.lds
@@ -5,6 +5,7 @@ SECTIONS
    . = 0xfffffc0000310000;
    _text = .;
    .text : { *(.text) }
+   .text2 : { *(.text2) }
    _etext = .;
 
   /* Exception table */
@@ -25,9 +26,12 @@ SECTIONS
   __init_begin = .;
   .text.init : { *(.text.init) }
   .data.init : { *(.data.init) }
-  . = ALIGN(8192);
+  . = ALIGN(2*8192);	/* Align double page for init_task_union */
   __init_end = .;
 
+  /* The initial task and kernel stack */
+  init_task : { *(init_task) }
+
   /* Global data */
   _data = .;
   .rodata : { *(.rodata) }
diff --git a/arch/i386/Makefile b/arch/i386/Makefile
index c67b8671a..bc4e03029 100644
--- a/arch/i386/Makefile
+++ b/arch/i386/Makefile
@@ -45,7 +45,7 @@ ifdef SMP
 CFLAGS := $(CFLAGS) -D__SMP__
 endif
 
-HEAD := arch/i386/kernel/head.o
+HEAD := arch/i386/kernel/head.o arch/i386/kernel/init_task.o
 
 SUBDIRS := $(SUBDIRS) arch/i386/kernel arch/i386/mm arch/i386/lib
 CORE_FILES := arch/i386/kernel/kernel.o arch/i386/mm/mm.o $(CORE_FILES)
@@ -85,7 +85,7 @@ bzdisk: vmlinux
 	@$(MAKEBOOT) BOOTIMAGE=bzImage zdisk
 
 install: vmlinux
-	@$(MAKEBOOT) install
+	@$(MAKEBOOT) BOOTIMAGE=bzImage install
 
 archclean:
 	@$(MAKEBOOT) clean
diff --git a/arch/i386/boot/video.S b/arch/i386/boot/video.S
index 8507b7081..d09f11283 100644
--- a/arch/i386/boot/video.S
+++ b/arch/i386/boot/video.S
@@ -1,12 +1,13 @@
 !
-!	Display adapter & video mode setup, version 2.10 (11-Nov-96)
+!	Display adapter & video mode setup, version 2.11 (03-May-97)
 !
-!	Copyright (C) 1995, 1996 Martin Mares <mj@k332.feld.cvut.cz>
+!	Copyright (C) 1995 -- 1997 Martin Mares <mj@k332.feld.cvut.cz>
 !	Based on the original setup.S code (C) Linus Torvalds and Mats Anderson
 !
 
-! Enable autodetection of SVGA adapters and modes
-#define CONFIG_VIDEO_SVGA
+! Enable autodetection of SVGA adapters and modes. If you really need this
+! feature, drop me a mail as I think of removing it some day...
+#undef CONFIG_VIDEO_SVGA
 
 ! Enable autodetection of VESA modes
 #define CONFIG_VIDEO_VESA
@@ -1807,11 +1808,11 @@ listhdr:	db	0x0d, 0x0a
 		.ascii	"Mode:    COLSxROWS:"
 crlft:		db	0x0d, 0x0a, 0
 prompt:		db	0x0d, 0x0a
-		.ascii	"Enter mode number: "
+		.ascii	"Enter mode number or `scan': "
 		db	0
 unknt:		.ascii	"Unknown mode ID. Try again."
 		db	0
-badmdt:		.ascii	"You passed an undefined mode number to setup."
+badmdt:		.ascii	"You passed an undefined mode number."
 		db	0x0d, 0x0a, 0
 vesaer:		.ascii	"Error: Scanning of VESA modes failed. Please "
 		.ascii	"report to <mj@k332.feld.cvut.cz>."
diff --git a/arch/i386/defconfig b/arch/i386/defconfig
index e5b5d9f84..a27b6ebce 100644
--- a/arch/i386/defconfig
+++ b/arch/i386/defconfig
@@ -141,6 +141,7 @@ CONFIG_SCSI_OMIT_FLASHPOINT=y
 # CONFIG_SCSI_QLOGIC_FAS is not set
 # CONFIG_SCSI_QLOGIC_ISP is not set
 # CONFIG_SCSI_SEAGATE is not set
+# CONFIG_SCSI_DC390T is not set
 # CONFIG_SCSI_T128 is not set
 # CONFIG_SCSI_U14_34F is not set
 # CONFIG_SCSI_ULTRASTOR is not set
diff --git a/arch/i386/kernel/Makefile b/arch/i386/kernel/Makefile
index e04fb5efb..9491ef562 100644
--- a/arch/i386/kernel/Makefile
+++ b/arch/i386/kernel/Makefile
@@ -15,7 +15,7 @@ else
 	$(CC) -D__ASSEMBLY__ -traditional -c $< -o $*.o
 endif
 
-all: kernel.o head.o
+all: kernel.o head.o init_task.o
 
 O_TARGET := kernel.o
 O_OBJS   := process.o signal.o entry.o traps.o irq.o vm86.o bios32.o \
diff --git a/arch/i386/kernel/bios32.c b/arch/i386/kernel/bios32.c
index e128000c3..157e62b2d 100644
--- a/arch/i386/kernel/bios32.c
+++ b/arch/i386/kernel/bios32.c
@@ -1,6 +1,8 @@
 /*
  * bios32.c - BIOS32, PCI BIOS functions.
  *
+ * $Id: bios32.c,v 1.11 1997/05/07 13:35:21 mj Exp $
+ *
  * Sponsored by
  *	iX Multiuser Multitasking Magazine
  *	Hannover, Germany
@@ -52,6 +54,11 @@
  * Feb 3, 1997  : Set internal functions to static, save/restore flags
  *	avoid dead locks reading broken PCI BIOS, werner@suse.de 
  *
+ * Apr 26, 1997 : Fixed case when there is BIOS32, but not PCI BIOS
+ *	(mj@atrey.karlin.mff.cuni.cz)
+ *
+ * May 7,  1997 : Added some missing cli()'s. [mj]
+ * 
  */
 
 #include <linux/config.h>
@@ -158,7 +165,7 @@ static unsigned long bios32_service(unsigned long service)
 	unsigned long entry;		/* %edx */
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%edi)"
 		: "=a" (return_code),
 		  "=b" (address),
@@ -173,10 +180,10 @@ static unsigned long bios32_service(unsigned long service)
 		case 0:
 			return address + entry;
 		case 0x80:	/* Not present */
-			printk("bios32_service(%ld) : not present\n", service);
+			printk("bios32_service(0x%lx) : not present\n", service);
 			return 0;
 		default: /* Shouldn't happen */
-			printk("bios32_service(%ld) : returned 0x%x, mail drew@colorado.edu\n",
+			printk("bios32_service(0x%lx) : returned 0x%x, mail drew@colorado.edu\n",
 				service, return_code);
 			return 0;
 	}
@@ -189,7 +196,7 @@ static struct {
 } pci_indirect = { 0, KERNEL_CS };
 
 
-__initfunc(static unsigned long check_pcibios(unsigned long memory_start, unsigned long memory_end))
+__initfunc(static int check_pcibios(void))
 {
 	unsigned long signature;
 	unsigned char present_status;
@@ -201,7 +208,7 @@ __initfunc(static unsigned long check_pcibios(unsigned long memory_start, unsign
 	if ((pcibios_entry = bios32_service(PCI_SERVICE))) {
 		pci_indirect.address = pcibios_entry | PAGE_OFFSET;
 
-		save_flags(flags);
+		save_flags(flags); cli();
 		__asm__("lcall (%%edi)\n\t"
 			"jc 1f\n\t"
 			"xor %%ah, %%ah\n"
@@ -212,7 +219,7 @@ __initfunc(static unsigned long check_pcibios(unsigned long memory_start, unsign
 			: "1" (PCIBIOS_PCI_BIOS_PRESENT),
 			  "D" (&pci_indirect)
 			: "bx", "cx");
-	restore_flags(flags);
+		restore_flags(flags);
 
 		present_status = (pack >> 16) & 0xff;
 		major_revision = (pack >> 8) & 0xff;
@@ -232,9 +239,10 @@ __initfunc(static unsigned long check_pcibios(unsigned long memory_start, unsign
 		if (pcibios_entry) {
 			printk ("pcibios_init : PCI BIOS revision %x.%02x entry at 0x%lx\n",
 				major_revision, minor_revision, pcibios_entry);
+			return 1;
 		}
 	}
-	return memory_start;
+	return 0;
 }
 
 
@@ -245,7 +253,7 @@ static int pci_bios_find_class (unsigned int class_code, unsigned short index,
 	unsigned long ret;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__ ("lcall (%%edi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -270,7 +278,7 @@ static int pci_bios_find_device (unsigned short vendor, unsigned short device_id
 	unsigned short ret;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%edi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -295,7 +303,7 @@ static int pci_bios_read_config_byte(unsigned char bus,
 	unsigned long bx = (bus << 8) | device_fn;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%esi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -317,7 +325,7 @@ static int pci_bios_read_config_word (unsigned char bus,
 	unsigned long bx = (bus << 8) | device_fn;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%esi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -339,7 +347,7 @@ static int pci_bios_read_config_dword (unsigned char bus,
 	unsigned long bx = (bus << 8) | device_fn;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%esi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -361,7 +369,7 @@ static int pci_bios_write_config_byte (unsigned char bus,
 	unsigned long bx = (bus << 8) | device_fn;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%esi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -383,7 +391,7 @@ static int pci_bios_write_config_word (unsigned char bus,
 	unsigned long bx = (bus << 8) | device_fn;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%esi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -405,7 +413,7 @@ static int pci_bios_write_config_dword (unsigned char bus,
 	unsigned long bx = (bus << 8) | device_fn;
 	unsigned long flags;
 
-	save_flags(flags);
+	save_flags(flags); cli();
 	__asm__("lcall (%%esi)\n\t"
 		"jc 1f\n\t"
 		"xor %%ah, %%ah\n"
@@ -476,7 +484,7 @@ static int pci_direct_find_class (unsigned int class_code, unsigned short index,
     struct pci_dev *dev;
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     for (dev = pci_devices; dev; dev = dev->next) {
 	if (dev->class == class_code) {
 	    if (curr == index) {
@@ -502,7 +510,7 @@ static int pci_conf1_read_config_byte(unsigned char bus, unsigned char device_fn
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
     switch (where & 3) {
 	case 0: *value = inb(0xCFC);
@@ -523,7 +531,7 @@ static int pci_conf1_read_config_word (unsigned char bus,
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);    
     if (where & 2)
     	*value = inw(0xCFE);
@@ -538,7 +546,7 @@ static int pci_conf1_read_config_dword (unsigned char bus, unsigned char device_
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
     *value = inl(0xCFC);
     restore_flags(flags);
@@ -550,7 +558,7 @@ static int pci_conf1_write_config_byte (unsigned char bus, unsigned char device_
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);    
     outb(value, 0xCFC);
     restore_flags(flags);
@@ -562,7 +570,7 @@ static int pci_conf1_write_config_word (unsigned char bus, unsigned char device_
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
     outw(value, 0xCFC);
     restore_flags(flags);
@@ -574,7 +582,7 @@ static int pci_conf1_write_config_dword (unsigned char bus, unsigned char device
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outl(CONFIG_CMD(bus,device_fn,where), 0xCF8);
     outl(value, 0xCFC);
     restore_flags(flags);
@@ -610,7 +618,7 @@ static int pci_conf2_read_config_byte(unsigned char bus, unsigned char device_fn
 
     if (device_fn & 0x80)
 	return PCIBIOS_DEVICE_NOT_FOUND;
-    save_flags(flags);
+    save_flags(flags); cli();
     outb (FUNC(device_fn), 0xCF8);
     outb (bus, 0xCFA);
     *value = inb(IOADDR(device_fn,where));
@@ -626,7 +634,7 @@ static int pci_conf2_read_config_word (unsigned char bus, unsigned char device_f
 
     if (device_fn & 0x80)
 	return PCIBIOS_DEVICE_NOT_FOUND;
-    save_flags(flags);
+    save_flags(flags); cli();
     outb (FUNC(device_fn), 0xCF8);
     outb (bus, 0xCFA);
     *value = inw(IOADDR(device_fn,where));
@@ -642,7 +650,7 @@ static int pci_conf2_read_config_dword (unsigned char bus, unsigned char device_
 
     if (device_fn & 0x80)
 	return PCIBIOS_DEVICE_NOT_FOUND;
-    save_flags(flags);
+    save_flags(flags); cli();
     outb (FUNC(device_fn), 0xCF8);
     outb (bus, 0xCFA);
     *value = inl (IOADDR(device_fn,where));    
@@ -656,7 +664,7 @@ static int pci_conf2_write_config_byte (unsigned char bus, unsigned char device_
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outb (FUNC(device_fn), 0xCF8);
     outb (bus, 0xCFA);
     outb (value, IOADDR(device_fn,where));
@@ -670,7 +678,7 @@ static int pci_conf2_write_config_word (unsigned char bus, unsigned char device_
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outb (FUNC(device_fn), 0xCF8);
     outb (bus, 0xCFA);
     outw (value, IOADDR(device_fn,where));
@@ -684,7 +692,7 @@ static int pci_conf2_write_config_dword (unsigned char bus, unsigned char device
 {
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
     outb (FUNC(device_fn), 0xCF8);
     outb (bus, 0xCFA);
     outl (value, IOADDR(device_fn,where));    
@@ -716,7 +724,7 @@ __initfunc(static struct pci_access *check_direct_pci(void))
     unsigned int tmp;
     unsigned long flags;
 
-    save_flags(flags);
+    save_flags(flags); cli();
 
     /*
      * check if configuration type 1 works
@@ -912,13 +920,11 @@ __initfunc(unsigned long pcibios_init(unsigned long memory_start, unsigned long
 				bios32_entry = check->fields.entry;
 				printk ("pcibios_init : BIOS32 Service Directory entry at 0x%lx\n", bios32_entry);
 				bios32_indirect.address = bios32_entry + PAGE_OFFSET;
- 				access_pci = &pci_bios_access;
 			}
 		}
 	}
-	if (bios32_entry) {
-		memory_start = check_pcibios (memory_start, memory_end);
-	}
+	if (bios32_entry && check_pcibios())
+ 		access_pci = &pci_bios_access;
 #endif
 	return memory_start;
 }
diff --git a/arch/i386/kernel/entry.S b/arch/i386/kernel/entry.S
index 84fe0c7fd..ac67da797 100644
--- a/arch/i386/kernel/entry.S
+++ b/arch/i386/kernel/entry.S
@@ -110,62 +110,45 @@ ENOSYS = 38
 	addl $4,%esp; \
 	iret
 
-#ifdef	__SMP__
-/* Get the processor ID multiplied by 4 */
-#define GET_PROCESSOR_OFFSET(reg) \
-	movl SYMBOL_NAME(apic_reg), reg; \
-	movl 32(reg), reg; \
-	shrl $22, reg; \
-	andl $0x3C, reg;
-
-#define GET_CURRENT(reg) \
-	GET_PROCESSOR_OFFSET(reg) \
-	movl SYMBOL_NAME(current_set)(reg),reg
-
-#else
-
 #define GET_CURRENT(reg) \
-	movl SYMBOL_NAME(current_set),reg
-
-#endif
+	movl %esp, reg; \
+	andl $-8192, reg;
 
 ENTRY(lcall7)
 	pushfl			# We get a different stack layout with call gates,
 	pushl %eax		# which has to be cleaned up later..
 	SAVE_ALL
-	GET_CURRENT(%ebx)
 	movl EIP(%esp),%eax	# due to call gates, this is eflags, not eip..
 	movl CS(%esp),%edx	# this is eip..
 	movl EFLAGS(%esp),%ecx	# and this is cs..
 	movl %eax,EFLAGS(%esp)	#
 	movl %edx,EIP(%esp)	# Now we move them to their "normal" places
 	movl %ecx,CS(%esp)	#
-	movl %esp,%eax
-	GET_CURRENT(%edx)
-	pushl %eax
-	movl exec_domain(%edx),%edx	# Get the execution domain
+	movl %esp,%ebx
+	pushl %ebx
+	andl $-8192,%ebx	# GET_CURRENT
+	movl exec_domain(%ebx),%edx	# Get the execution domain
 	movl 4(%edx),%edx	# Get the lcall7 handler for the domain
 	call *%edx
 	popl %eax
 	jmp ret_from_sys_call
 
+
 #ifdef __SMP__
 	ALIGN
 	.globl	ret_from_smpfork
 ret_from_smpfork:
+	GET_CURRENT(%ebx)
 	btrl	$0, SYMBOL_NAME(scheduler_lock)
 	jmp	ret_from_sys_call
 #endif /* __SMP__ */
 
-	ALIGN
-handle_bottom_half:
-	pushl $2f
-	jmp SYMBOL_NAME(do_bottom_half)
-
-	ALIGN
-reschedule:
-	pushl $ret_from_sys_call
-	jmp SYMBOL_NAME(schedule)    # test
+/*
+ * Return to user mode is not as complex as all this looks,
+ * but we want the default path for a system call return to
+ * go as quickly as possible which is why some of this is
+ * less clear than it otherwise should be.
+ */
 
 ENTRY(system_call)
 	pushl %eax			# save orig_eax
@@ -180,16 +163,11 @@ ENTRY(system_call)
 	ALIGN
 	.globl ret_from_sys_call
 	.globl ret_from_intr
-ret_from_intr:
 ret_from_sys_call:
-	GET_CURRENT(%ebx)
 	movl SYMBOL_NAME(bh_mask),%eax
 	andl SYMBOL_NAME(bh_active),%eax
 	jne handle_bottom_half
-2:	movl EFLAGS(%esp),%eax		# mix EFLAGS and CS
-	movb CS(%esp),%al
-	testl $(VM_MASK | 3),%eax	# return to VM86 mode or non-supervisor?
-	je 1f
+ret_with_reschedule:
 	cmpl $0,SYMBOL_NAME(need_resched)
 	jne reschedule
 	movl blocked(%ebx),%eax
@@ -197,7 +175,6 @@ ret_from_sys_call:
 	notl %eax
 	andl signal(%ebx),%eax
 	jne signal_return
-1:	
 	RESTORE_ALL
 	ALIGN
 signal_return:
@@ -230,6 +207,30 @@ badsys:
 	movl $-ENOSYS,EAX(%esp)
 	jmp ret_from_sys_call
 
+	ALIGN
+ret_from_exception:
+	movl SYMBOL_NAME(bh_mask),%eax
+	andl SYMBOL_NAME(bh_active),%eax
+	jne handle_bottom_half
+	ALIGN
+ret_from_intr:
+	GET_CURRENT(%ebx)
+	movl EFLAGS(%esp),%eax		# mix EFLAGS and CS
+	movb CS(%esp),%al
+	testl $(VM_MASK | 3),%eax	# return to VM86 mode or non-supervisor?
+	jne ret_with_reschedule
+	RESTORE_ALL
+
+	ALIGN
+handle_bottom_half:
+	pushl $ret_from_intr
+	jmp SYMBOL_NAME(do_bottom_half)
+
+	ALIGN
+reschedule:
+	pushl $ret_from_sys_call
+	jmp SYMBOL_NAME(schedule)    # test
+
 
 ENTRY(divide_error)
 	pushl $0		# no error code
@@ -260,7 +261,7 @@ error_code:
 	GET_CURRENT(%ebx)
 	call *%ecx
 	addl $8,%esp
-	jmp ret_from_sys_call
+	jmp ret_from_exception
 
 ENTRY(coprocessor_error)
 	pushl $0
@@ -271,7 +272,7 @@ ENTRY(device_not_available)
 	pushl $-1		# mark this as an int
 	SAVE_ALL
 	GET_CURRENT(%ebx)
-	pushl $ret_from_sys_call
+	pushl $ret_from_exception
 	movl %cr0,%eax
 	testl $0x4,%eax			# EM (math emulation bit)
 	je SYMBOL_NAME(math_state_restore)
diff --git a/arch/i386/kernel/head.S b/arch/i386/kernel/head.S
index 2bd095997..a42b87b1b 100644
--- a/arch/i386/kernel/head.S
+++ b/arch/i386/kernel/head.S
@@ -39,19 +39,21 @@ startup_32:
 	jz 1f
 /*
  *	New page tables may be in 4Mbyte page mode and may
- *	be using the global pages.
+ *	be using the global pages. 
+ *
+ *	NOTE! We have to correct for the fact that we're
+ *	not yet offset 0xC0000000..
  */
+#define cr4_bits mmu_cr4_features-0xC0000000
 #ifdef GAS_KNOWS_CR4
 	movl %cr4,%eax		# Turn on 4Mb pages
-	orl $16+128,%eax
+	orl cr4_bits,%eax
 	movl %eax,%cr4
 #else
 	.byte 0x0f,0x20,0xe0
-	orl $16+128,%eax
+	orl cr4_bits,%eax
 	.byte 0x0f,0x22,0xe0
 #endif
-	movl %eax,%cr3		/* flush TLB as per app note */
-	movl %cr0,%eax
 #endif
 /*
  * Setup paging (the tables are already set up, just switch them on)
@@ -67,24 +69,16 @@ startup_32:
 	movl $1f,%eax
 	jmp *%eax		/* make sure eip is relocated */
 1:
+	/* Set up the stack pointer */
+	lss stack_start,%esp
 
 #ifdef __SMP__
 	orw  %bx,%bx
 	jz  1f				/* Initial CPU cleans BSS */
-/*
- *	Set up the stack
- */
-        movl $(KERNEL_DS),%eax         /* walken modif */
-	mov %ax,%ss
-	xorl %eax,%eax
-	movw %cx, %ax
-	movl %eax,%esp
-	addl $0xC0000000, %esp		/* shift it to the upper mapping */
 	pushl $0
 	popfl
 	jmp checkCPUtype
 1:
-	lss stack_start,%esp
 #endif __SMP__
 /*
  * Clear BSS first so that there are no surprises...
@@ -305,15 +299,53 @@ rp_sidt:
 	jne rp_sidt
 	ret
 
+ENTRY(stack_start)
+	.long SYMBOL_NAME(init_task_union)+8192
+	.long KERNEL_DS
+
+/* This is the default interrupt "handler" :-) */
+int_msg:
+	.asciz "Unknown interrupt\n"
+	ALIGN
+ignore_int:
+	cld
+	pushl %eax
+	pushl %ecx
+	pushl %edx
+	push %ds
+	movl $(KERNEL_DS),%eax
+	mov %ax,%ds
+	mov %ax,%es
+	mov %ax,%fs
+	pushl $int_msg
+	call SYMBOL_NAME(printk)
+	popl %eax
+	pop %ds
+	popl %edx
+	popl %ecx
+	popl %eax
+	iret
+
+/*
+ * The interrupt descriptor table has room for 256 idt's
+ */
+	ALIGN
+.word 0
+idt_descr:
+	.word 256*8-1		# idt contains 256 entries
+	.long SYMBOL_NAME(idt)
+
+	ALIGN
+.word 0
+gdt_descr:
+#ifdef CONFIG_APM
+	.word (11+2*NR_TASKS)*8-1
+#else
+	.word (8+2*NR_TASKS)*8-1
+#endif
+	.long SYMBOL_NAME(gdt)
 
 /*
- * page 0 is made non-existent, so that kernel NULL pointer references get
- * caught. Thus the swapper page directory has been moved to 0x101000
- * with the introduction of the compressed boot code.  Theoretically,
- * the original design of overlaying the startup code with the swapper
- * page directory is still possible --- it would reduce the size of the kernel
- * by 2-3k.  This would be a good thing to do at some point.....
- *
  * This is initialized to create a identity-mapping at 0-4M (for bootup
  * purposes) and another mapping of the 0-4M area at virtual address
  * 0xC0000000.
@@ -471,63 +503,29 @@ ENTRY(empty_bad_page_table)
 ENTRY(empty_zero_page)
 
 .org 0x6000
-
-stack_start:
-	.long SYMBOL_NAME(init_user_stack)+4096
-	.long KERNEL_DS
-
-/* This is the default interrupt "handler" :-) */
-int_msg:
-	.asciz "Unknown interrupt\n"
-	ALIGN
-ignore_int:
-	cld
-	pushl %eax
-	pushl %ecx
-	pushl %edx
-	push %ds
-	push %es
-	push %fs
-	movl $(KERNEL_DS),%eax
-	mov %ax,%ds
-	mov %ax,%es
-	mov %ax,%fs
-	pushl $int_msg
-	call SYMBOL_NAME(printk)
-	popl %eax
-	pop %fs
-	pop %es
-	pop %ds
-	popl %edx
-	popl %ecx
-	popl %eax
-	iret
+ENTRY(this_must_match_init_task)
 
 /*
- * The interrupt descriptor table has room for 256 idt's
+ * This starts the data section. Note that the above is all
+ * in the text section because it has alignment requirements
+ * that we cannot fulfill any other way.
  */
-	ALIGN
-.word 0
-idt_descr:
-	.word 256*8-1		# idt contains 256 entries
-	.long SYMBOL_NAME(idt)
+.data
 
+ALIGN
+/* 256 quadwords - 2048 bytes of idt */
 ENTRY(idt)
 	.fill 256,8,0		# idt is uninitialized
 
-	ALIGN
-.word 0
-gdt_descr:
-#ifdef CONFIG_APM
-	.word (11+2*NR_TASKS)*8-1
-#else
-	.word (8+2*NR_TASKS)*8-1
-#endif
-	.long SYMBOL_NAME(gdt)
-
 /*
  * This gdt setup gives the kernel a 1GB address space at virtual
  * address 0xC0000000 - space enough for expansion, I hope.
+ *
+ * This contains up to 8192 quadwords depending on NR_TASKS - 64kB of
+ * gdt entries.  Ugh. 
+ *
+ * NOTE! Make sure the gdt descriptor in head.S matches this if you
+ * change anything.
  */
 ENTRY(gdt)
 	.quad 0x0000000000000000	/* NULL descriptor */
diff --git a/arch/i386/kernel/i386_ksyms.c b/arch/i386/kernel/i386_ksyms.c
index 8c16f0204..daa6baf42 100644
--- a/arch/i386/kernel/i386_ksyms.c
+++ b/arch/i386/kernel/i386_ksyms.c
@@ -19,6 +19,11 @@ extern void dump_thread(struct pt_regs *, struct user *);
 extern int dump_fpu(elf_fpregset_t *);
 extern void __lock_kernel(void);
 
+#if defined(CONFIG_BLK_DEV_IDE) || defined(CONFIG_BLK_DEV_HD) || defined(CONFIG_BLK_DEV_IDE_MODULE) || defined(CONFIG_BLK_DEV_HD_MODULE)
+extern struct drive_info_struct drive_info;
+EXPORT_SYMBOL(drive_info);
+#endif
+
 /* platform dependent support */
 EXPORT_SYMBOL(EISA_bus);
 EXPORT_SYMBOL(MCA_bus);
@@ -39,12 +44,13 @@ EXPORT_SYMBOL(csum_partial_copy);
 #ifdef __SMP__
 EXPORT_SYMBOL(apic_reg);	/* Needed internally for the I386 inlines */
 EXPORT_SYMBOL(cpu_data);
-EXPORT_SYMBOL(kernel_flag);
-EXPORT_SYMBOL(active_kernel_processor);
+EXPORT_SYMBOL_NOVERS(kernel_flag);
+EXPORT_SYMBOL_NOVERS(active_kernel_processor);
 EXPORT_SYMBOL(smp_invalidate_needed);
 EXPORT_SYMBOL_NOVERS(__lock_kernel);
 
 /* Global SMP irq stuff */
+EXPORT_SYMBOL(synchronize_irq);
 EXPORT_SYMBOL(global_irq_holder);
 EXPORT_SYMBOL(__global_cli);
 EXPORT_SYMBOL(__global_sti);
diff --git a/arch/i386/kernel/init_task.c b/arch/i386/kernel/init_task.c
new file mode 100644
index 000000000..cc0a19231
--- /dev/null
+++ b/arch/i386/kernel/init_task.c
@@ -0,0 +1,22 @@
+#include <linux/mm.h>
+#include <linux/sched.h>
+
+#include <asm/pgtable.h>
+
+static struct vm_area_struct init_mmap = INIT_MMAP;
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS;
+struct mm_struct init_mm = INIT_MM;
+
+/*
+ * Initial task structure.
+ *
+ * We need to make sure that this is 8192-byte aligned due to the
+ * way process stacks are handled. This is done by making sure
+ * the linker maps this in the .text segment right after head.S,
+ * and making head.S ensure the proper alignment.
+ *
+ * The things we do for performance..
+ */
+union task_union init_task_union __attribute__((__section__(".text"))) = { INIT_TASK };
diff --git a/arch/i386/kernel/irq.c b/arch/i386/kernel/irq.c
index 95a7b525f..e5fb5acb1 100644
--- a/arch/i386/kernel/irq.c
+++ b/arch/i386/kernel/irq.c
@@ -44,9 +44,6 @@ extern volatile unsigned long smp_local_timer_ticks[1+NR_CPUS];
 
 #define CR0_NE 32
 
-static unsigned char cache_21 = 0xff;
-static unsigned char cache_A1 = 0xff;
-
 unsigned int local_irq_count[NR_CPUS];
 #ifdef __SMP__
 atomic_t __intel_bh_counter;
@@ -58,51 +55,84 @@ int __intel_bh_counter;
 static unsigned int int_count[NR_CPUS][NR_IRQS] = {{0},};
 #endif
 
-static inline void mask_irq(unsigned int irq_nr)
-{
-	unsigned char mask;
+/*
+ * This contains the irq mask for both irq controllers
+ */
+static unsigned int cached_irq_mask = 0xffff;
+
+#define cached_21	(((char *)(&cached_irq_mask))[0])
+#define cached_A1	(((char *)(&cached_irq_mask))[1])
 
-	mask = 1 << (irq_nr & 7);
-	if (irq_nr < 8) {
-		cache_21 |= mask;
-		outb(cache_21,0x21);
+spinlock_t irq_controller_lock;
+
+/*
+ * This is always called from an interrupt context
+ * with local interrupts disabled. Don't worry about
+ * irq-safe locks.
+ *
+ * Note that we always ack the primary irq controller,
+ * even if the interrupt came from the secondary, as
+ * the primary will still have routed it. Oh, the joys
+ * of PC hardware.
+ */
+static inline void mask_and_ack_irq(int irq_nr)
+{
+	spin_lock(&irq_controller_lock);
+	cached_irq_mask |= 1 << irq_nr;
+	if (irq_nr & 8) {
+		inb(0xA1);	/* DUMMY */
+		outb(cached_A1,0xA1);
+		outb(0x20,0xA0);
 	} else {
-		cache_A1 |= mask;
-		outb(cache_A1,0xA1);
+		inb(0x21);	/* DUMMY */
+		outb(cached_21,0x21);
 	}
+	outb(0x20,0x20);
+	spin_unlock(&irq_controller_lock);
 }
 
-static inline void unmask_irq(unsigned int irq_nr)
+static inline void set_irq_mask(int irq_nr)
 {
-	unsigned char mask;
-
-	mask = ~(1 << (irq_nr & 7));
-	if (irq_nr < 8) {
-		cache_21 &= mask;
-		outb(cache_21,0x21);
+	if (irq_nr & 8) {
+		outb(cached_A1,0xA1);
 	} else {
-		cache_A1 &= mask;
-		outb(cache_A1,0xA1);
+		outb(cached_21,0x21);
 	}
 }
 
+/*
+ * These have to be protected by the spinlock
+ * before being called.
+ */
+static inline void mask_irq(unsigned int irq_nr)
+{
+	cached_irq_mask |= 1 << irq_nr;
+	set_irq_mask(irq_nr);
+}
+
+static inline void unmask_irq(unsigned int irq_nr)
+{
+	cached_irq_mask &= ~(1 << irq_nr);
+	set_irq_mask(irq_nr);
+}
+
 void disable_irq(unsigned int irq_nr)
 {
 	unsigned long flags;
 
-	save_flags(flags);
-	cli();
+	spin_lock_irqsave(&irq_controller_lock, flags);
 	mask_irq(irq_nr);
-	restore_flags(flags);
+	spin_unlock_irqrestore(&irq_controller_lock, flags);
+	synchronize_irq();
 }
 
 void enable_irq(unsigned int irq_nr)
 {
 	unsigned long flags;
-	save_flags(flags);
-	cli();
+
+	spin_lock_irqsave(&irq_controller_lock, flags);
 	unmask_irq(irq_nr);
-	restore_flags(flags);
+	spin_unlock_irqrestore(&irq_controller_lock, flags);
 }
 
 /*
@@ -133,7 +163,8 @@ void enable_irq(unsigned int irq_nr)
 #error make irq stub building NR_IRQS dependent and remove me.
 #endif
 
-BUILD_TIMER_IRQ(FIRST,0,0x01)
+BUILD_COMMON_IRQ()
+BUILD_IRQ(FIRST,0,0x01)
 BUILD_IRQ(FIRST,1,0x02)
 BUILD_IRQ(FIRST,2,0x04)
 BUILD_IRQ(FIRST,3,0x08)
@@ -157,10 +188,6 @@ BUILD_SMP_INTERRUPT(stop_cpu_interrupt)
 BUILD_SMP_TIMER_INTERRUPT(apic_timer_interrupt)
 #endif
 
-/*
- * Pointers to the low-level handlers: first the general ones, then the
- * fast ones, then the bad ones.
- */
 static void (*interrupt[17])(void) = {
 	IRQ0_interrupt, IRQ1_interrupt, IRQ2_interrupt, IRQ3_interrupt,
 	IRQ4_interrupt, IRQ5_interrupt, IRQ6_interrupt, IRQ7_interrupt,
@@ -168,28 +195,6 @@ static void (*interrupt[17])(void) = {
 	IRQ12_interrupt, IRQ13_interrupt, IRQ14_interrupt, IRQ15_interrupt	
 };
 
-static void (*fast_interrupt[16])(void) = {
-	fast_IRQ0_interrupt, fast_IRQ1_interrupt,
-	fast_IRQ2_interrupt, fast_IRQ3_interrupt,
-	fast_IRQ4_interrupt, fast_IRQ5_interrupt,
-	fast_IRQ6_interrupt, fast_IRQ7_interrupt,
-	fast_IRQ8_interrupt, fast_IRQ9_interrupt,
-	fast_IRQ10_interrupt, fast_IRQ11_interrupt,
-	fast_IRQ12_interrupt, fast_IRQ13_interrupt,
-	fast_IRQ14_interrupt, fast_IRQ15_interrupt
-};
-
-static void (*bad_interrupt[16])(void) = {
-	bad_IRQ0_interrupt, bad_IRQ1_interrupt,
-	bad_IRQ2_interrupt, bad_IRQ3_interrupt,
-	bad_IRQ4_interrupt, bad_IRQ5_interrupt,
-	bad_IRQ6_interrupt, bad_IRQ7_interrupt,
-	bad_IRQ8_interrupt, bad_IRQ9_interrupt,
-	bad_IRQ10_interrupt, bad_IRQ11_interrupt,
-	bad_IRQ12_interrupt, bad_IRQ13_interrupt,
-	bad_IRQ14_interrupt, bad_IRQ15_interrupt
-};
-
 /*
  * Initial irq handlers.
  */
@@ -240,14 +245,10 @@ int get_irq_list(char *buf)
 		action = irq_action[i];
 		if (!action) 
 			continue;
-		len += sprintf(buf+len, "%2d: %10u %c %s",
-			i, kstat.interrupts[i],
-			(action->flags & SA_INTERRUPT) ? '+' : ' ',
-			action->name);
+		len += sprintf(buf+len, "%2d: %10u   %s",
+			i, kstat.interrupts[i], action->name);
 		for (action=action->next; action; action = action->next) {
-			len += sprintf(buf+len, ",%s %s",
-				(action->flags & SA_INTERRUPT) ? " +" : "",
-				action->name);
+			len += sprintf(buf+len, ", %s", action->name);
 		}
 		len += sprintf(buf+len, "\n");
 	}
@@ -298,13 +299,9 @@ int get_smp_prof_list(char *buf) {
 		for (j=0;j<smp_num_cpus;j++)
 			len+=sprintf(buf+len, "%10d ",
 				int_count[cpu_logical_map[j]][i]);
-		len += sprintf(buf+len, "%c %s",
-			(action->flags & SA_INTERRUPT) ? '+' : ' ',
-			action->name);
+		len += sprintf(buf+len, "  %s", action->name);
 		for (action=action->next; action; action = action->next) {
-			len += sprintf(buf+len, ",%s %s",
-				(action->flags & SA_INTERRUPT) ? " +" : "",
-				action->name);
+			len += sprintf(buf+len, ", %s", action->name);
 		}
 		len += sprintf(buf+len, "\n");
 	}
@@ -393,16 +390,8 @@ static inline void check_smp_invalidate(int cpu)
 
 static unsigned long previous_irqholder;
 
-#undef INIT_STUCK
-#define INIT_STUCK 100000000
-
-#undef STUCK
-#define STUCK \
-if (!--stuck) {printk("wait_on_irq CPU#%d stuck at %08lx, waiting for %08lx (local=%d, global=%d)\n", cpu, where, previous_irqholder, local_count, atomic_read(&global_irq_count)); stuck = INIT_STUCK; }
-
 static inline void wait_on_irq(int cpu, unsigned long where)
 {
-	int stuck = INIT_STUCK;
 	int local_count = local_irq_count[cpu];
 
 	/* Are we the only one in an interrupt context? */
@@ -421,13 +410,12 @@ static inline void wait_on_irq(int cpu, unsigned long where)
 		 * their things before trying to get the lock again.
 		 */
 		for (;;) {
-			STUCK;
 			check_smp_invalidate(cpu);
 			if (atomic_read(&global_irq_count))
 				continue;
 			if (global_irq_lock)
 				continue;
-			if (!set_bit(0,&global_irq_lock))
+			if (!test_and_set_bit(0,&global_irq_lock))
 				break;
 		}
 		atomic_add(local_count, &global_irq_count);
@@ -456,28 +444,18 @@ void synchronize_irq(void)
 	}
 }
 
-#undef INIT_STUCK
-#define INIT_STUCK 10000000
-
-#undef STUCK
-#define STUCK \
-if (!--stuck) {printk("get_irqlock stuck at %08lx, waiting for %08lx\n", where, previous_irqholder); stuck = INIT_STUCK;}
-
 static inline void get_irqlock(int cpu, unsigned long where)
 {
-	int stuck = INIT_STUCK;
-
-	if (set_bit(0,&global_irq_lock)) {
+	if (test_and_set_bit(0,&global_irq_lock)) {
 		/* do we already hold the lock? */
 		if ((unsigned char) cpu == global_irq_holder)
 			return;
 		/* Uhhuh.. Somebody else got it. Wait.. */
 		do {
 			do {
-				STUCK;
 				check_smp_invalidate(cpu);
 			} while (test_bit(0,&global_irq_lock));
-		} while (set_bit(0,&global_irq_lock));		
+		} while (test_and_set_bit(0,&global_irq_lock));		
 	}
 	/*
 	 * Ok, we got the lock bit.
@@ -519,7 +497,8 @@ void __global_restore_flags(unsigned long flags)
 {
 	switch (flags) {
 	case 0:
-		__global_sti();
+		release_irqlock(smp_processor_id());
+		__sti();
 		break;
 	case 1:
 		__global_cli();
@@ -533,56 +512,58 @@ void __global_restore_flags(unsigned long flags)
 #endif
 
 /*
- * do_IRQ handles IRQ's that have been installed without the
- * SA_INTERRUPT flag: it uses the full signal-handling return
- * and runs with other interrupts enabled. All relatively slow
- * IRQ's should use this format: notably the keyboard/timer
- * routines.
+ * do_IRQ handles all normal device IRQ's (the special
+ * SMP cross-CPU interrupts have their own specific
+ * handlers).
  */
-asmlinkage void do_IRQ(int irq, struct pt_regs * regs)
+asmlinkage void do_IRQ(struct pt_regs regs)
 {
+	int irq = regs.orig_eax & 0xff;
 	struct irqaction * action;
-	int do_random, cpu = smp_processor_id();
+	int status, cpu;
+
+	/* 
+	 * mask and ack quickly, we don't want the irq controller
+	 * thinking we're snobs just because some other CPU has
+	 * disabled global interrupts (we have already done the
+	 * INT_ACK cycles, it's too late to try to pretend to the
+	 * controller that we aren't taking the interrupt).
+	 */
+	mask_and_ack_irq(irq);
 
+	cpu = smp_processor_id();
 	irq_enter(cpu, irq);
 	kstat.interrupts[irq]++;
 
-	/* slow interrupts run with interrupts enabled */
-	__sti();
+	/* Return with this interrupt masked if no action */
+	status = 0;
 	action = *(irq + irq_action);
-	do_random = 0;
-	while (action) {
-		do_random |= action->flags;
-		action->handler(irq, action->dev_id, regs);
-		action = action->next;
+	if (action) {
+		do {
+			status |= action->flags;
+			action->handler(irq, action->dev_id, &regs);
+			action = action->next;
+		} while (action);
+		if (status & SA_SAMPLE_RANDOM)
+			add_interrupt_randomness(irq);
+
+		__cli();
+		spin_lock(&irq_controller_lock);
+		unmask_irq(irq);
+		spin_unlock(&irq_controller_lock);
 	}
-	if (do_random & SA_SAMPLE_RANDOM)
-		add_interrupt_randomness(irq);
-	irq_exit(cpu, irq);
-}
 
-/*
- * do_fast_IRQ handles IRQ's that don't need the fancy interrupt return
- * stuff - the handler is also running with interrupts disabled unless
- * it explicitly enables them later.
- */
-asmlinkage void do_fast_IRQ(int irq)
-{
-	struct irqaction * action;
-	int do_random, cpu = smp_processor_id();
-
-	irq_enter(cpu, irq);
-	kstat.interrupts[irq]++;
-	action = *(irq + irq_action);
-	do_random = 0;
-	while (action) {
-		do_random |= action->flags;
-		action->handler(irq, action->dev_id, NULL);
-		action = action->next;
-	}
-	if (do_random & SA_SAMPLE_RANDOM)
-		add_interrupt_randomness(irq);
 	irq_exit(cpu, irq);
+	/*
+	 * This should be conditional: we should really get
+	 * a return code from the irq handler to tell us
+	 * whether the handler wants us to do software bottom
+	 * half handling or not..
+	 */
+	if (1) {
+		if (bh_active & bh_mask)
+			do_bottom_half();
+	}
 }
 
 int setup_x86_irq(int irq, struct irqaction * new)
@@ -597,10 +578,6 @@ int setup_x86_irq(int irq, struct irqaction * new)
 		if (!(old->flags & new->flags & SA_SHIRQ))
 			return -EBUSY;
 
-		/* Can't share interrupts unless both are same type */
-		if ((old->flags ^ new->flags) & SA_INTERRUPT)
-			return -EBUSY;
-
 		/* add new interrupt at end of irq queue */
 		do {
 			p = &old->next;
@@ -617,11 +594,9 @@ int setup_x86_irq(int irq, struct irqaction * new)
 	*p = new;
 
 	if (!shared) {
-		if (new->flags & SA_INTERRUPT)
-			set_intr_gate(0x20+irq,fast_interrupt[irq]);
-		else
-			set_intr_gate(0x20+irq,interrupt[irq]);
+		spin_lock(&irq_controller_lock);
 		unmask_irq(irq);
+		spin_unlock(&irq_controller_lock);
 	}
 	restore_flags(flags);
 	return 0;
@@ -676,10 +651,6 @@ void free_irq(unsigned int irq, void *dev_id)
 		save_flags(flags);
 		cli();
 		*p = action->next;
-		if (!irq[irq_action]) {
-			mask_irq(irq);
-			set_intr_gate(0x20+irq,bad_interrupt[irq]);
-		}
 		restore_flags(flags);
 		kfree(action);
 		return;
@@ -689,7 +660,7 @@ void free_irq(unsigned int irq, void *dev_id)
 
 unsigned long probe_irq_on (void)
 {
-	unsigned int i, irqs = 0, irqmask;
+	unsigned int i, irqs = 0;
 	unsigned long delay;
 
 	/* first, enable any unassigned irqs */
@@ -705,19 +676,17 @@ unsigned long probe_irq_on (void)
 		/* about 100ms delay */;
 
 	/* now filter out any obviously spurious interrupts */
-	irqmask = (((unsigned int)cache_A1)<<8) | (unsigned int)cache_21;
-	return irqs & ~irqmask;
+	return irqs & ~cached_irq_mask;
 }
 
 int probe_irq_off (unsigned long irqs)
 {
-	unsigned int i, irqmask;
+	unsigned int i;
 
-	irqmask = (((unsigned int)cache_A1)<<8) | (unsigned int)cache_21;
 #ifdef DEBUG
-	printk("probe_irq_off: irqs=0x%04lx irqmask=0x%04x\n", irqs, irqmask);
+	printk("probe_irq_off: irqs=0x%04lx irqmask=0x%04x\n", irqs, cached_irq_mask);
 #endif
-	irqs &= irqmask;
+	irqs &= cached_irq_mask;
 	if (!irqs)
 		return 0;
 	i = ffz(~irqs);
@@ -729,10 +698,6 @@ int probe_irq_off (unsigned long irqs)
 __initfunc(void init_IRQ(void))
 {
 	int i;
-	static unsigned char smptrap=0;
-	if(smptrap)
-		return;
-	smptrap=1;
 
 	/* set the clock to 100 Hz */
 	outb_p(0x34,0x43);		/* binary, mode 2, LSB/MSB, ch 0 */
@@ -740,7 +705,7 @@ __initfunc(void init_IRQ(void))
 	outb(LATCH >> 8 , 0x40);	/* MSB */
 
 	for (i = 0; i < NR_IRQS ; i++)
-		set_intr_gate(0x20+i,bad_interrupt[i]);
+		set_intr_gate(0x20+i,interrupt[i]);
 
 #ifdef __SMP__	
 	/*
diff --git a/arch/i386/kernel/irq.h b/arch/i386/kernel/irq.h
index 3a349f20a..1f9e89399 100644
--- a/arch/i386/kernel/irq.h
+++ b/arch/i386/kernel/irq.h
@@ -33,7 +33,6 @@ static inline void irq_enter(int cpu, int irq)
 
 static inline void irq_exit(int cpu, int irq)
 {
-	__cli();
 	hardirq_exit(cpu);
 	release_irqlock(cpu);
 }
@@ -63,125 +62,12 @@ static inline void irq_exit(int cpu, int irq)
 	"mov %dx,%ds\n\t" \
 	"mov %dx,%es\n\t"
 
-/*
- * SAVE_MOST/RESTORE_MOST is used for the faster version of IRQ handlers,
- * installed by using the SA_INTERRUPT flag. These kinds of IRQ's don't
- * call the routines that do signal handling etc on return, and can have
- * more relaxed register-saving etc. They are also atomic, and are thus
- * suited for small, fast interrupts like the serial lines or the harddisk
- * drivers, which don't actually need signal handling etc.
- *
- * Also note that we actually save only those registers that are used in
- * C subroutines (%eax, %edx and %ecx), so if you do something weird,
- * you're on your own. The only segments that are saved (not counting the
- * automatic stack and code segment handling) are %ds and %es, and they
- * point to kernel space. No messing around with %fs here.
- */
-#define SAVE_MOST \
-	"cld\n\t" \
-	"push %es\n\t" \
-	"push %ds\n\t" \
-	"pushl %eax\n\t" \
-	"pushl %edx\n\t" \
-	"pushl %ecx\n\t" \
-	"movl $" STR(KERNEL_DS) ",%edx\n\t" \
-	"mov %dx,%ds\n\t" \
-	"mov %dx,%es\n\t"
-
-#define RESTORE_MOST \
-	"popl %ecx\n\t" \
-	"popl %edx\n\t" \
-	"popl %eax\n\t" \
-	"pop %ds\n\t" \
-	"pop %es\n\t" \
-	"iret"
-
-/*
- * Some fast irq handlers might want to access saved registers (mostly
- * cs or flags)
- */
-
-struct fast_irq_regs {
-	long ecx;
-	long edx;
-	long eax;
-	int  xds;
-	int  xes;
-	long eip;
-	int  xcs;
-	long eflags;
-	long esp;
-	int  xss;
-};
-
-/*
- * The "inb" instructions are not needed, but seem to change the timings
- * a bit - without them it seems that the harddisk driver won't work on
- * all hardware. Arghh.
- */
-#define ACK_FIRST(mask,nr) \
-	"inb $0x21,%al\n\t" \
-	"jmp 1f\n" \
-	"1:\tjmp 1f\n" \
-	"1:\torb $" #mask ","SYMBOL_NAME_STR(cache_21)"\n\t" \
-	"movb "SYMBOL_NAME_STR(cache_21)",%al\n\t" \
-	"outb %al,$0x21\n\t" \
-	"jmp 1f\n" \
-	"1:\tjmp 1f\n" \
-	"1:\tmovb $0x20,%al\n\t" \
-	"outb %al,$0x20\n\t"
-
-#define ACK_SECOND(mask,nr) \
-	"inb $0xA1,%al\n\t" \
-	"jmp 1f\n" \
-	"1:\tjmp 1f\n" \
-	"1:\torb $" #mask ","SYMBOL_NAME_STR(cache_A1)"\n\t" \
-	"movb "SYMBOL_NAME_STR(cache_A1)",%al\n\t" \
-	"outb %al,$0xA1\n\t" \
-	"jmp 1f\n" \
-	"1:\tjmp 1f\n" \
-	"1:\tmovb $0x20,%al\n\t" \
-	"outb %al,$0xA0\n\t" \
-	"jmp 1f\n" \
-	"1:\tjmp 1f\n" \
-	"1:\toutb %al,$0x20\n\t"
-
-#define UNBLK_FIRST(mask) \
-	"inb $0x21,%al\n\t" \
-	"jmp 1f\n" \
-	"1:\tjmp 1f\n" \
-	"1:\tandb $~(" #mask "),"SYMBOL_NAME_STR(cache_21)"\n\t" \
-	"movb "SYMBOL_NAME_STR(cache_21)",%al\n\t" \
-	"outb %al,$0x21\n\t"
-
-#define UNBLK_SECOND(mask) \
-	"inb $0xA1,%al\n\t" \
-	"jmp 1f\n" \
-	"1:\tjmp 1f\n" \
-	"1:\tandb $~(" #mask "),"SYMBOL_NAME_STR(cache_A1)"\n\t" \
-	"movb "SYMBOL_NAME_STR(cache_A1)",%al\n\t" \
-	"outb %al,$0xA1\n\t"
-
 #define IRQ_NAME2(nr) nr##_interrupt(void)
 #define IRQ_NAME(nr) IRQ_NAME2(IRQ##nr)
-#define FAST_IRQ_NAME(nr) IRQ_NAME2(fast_IRQ##nr)
-#define BAD_IRQ_NAME(nr) IRQ_NAME2(bad_IRQ##nr)
-
-#ifdef	__SMP__
-
-#define GET_CURRENT \
-	"movl "SYMBOL_NAME_STR(apic_reg)", %ebx\n\t" \
-	"movl 32(%ebx), %ebx\n\t" \
-	"shrl $22,%ebx\n\t" \
-	"andl $0x3C,%ebx\n\t" \
-	"movl " SYMBOL_NAME_STR(current_set) "(,%ebx),%ebx\n\t"
-	
-#else
 
 #define GET_CURRENT \
-	"movl " SYMBOL_NAME_STR(current_set) ",%ebx\n\t"
-	
-#endif
+	"movl %esp, %ebx\n\t" \
+	"andl $-8192, %ebx\n\t"
 
 #ifdef __SMP__
 
@@ -205,66 +91,30 @@ __asm__( \
 "\n"__ALIGN_STR"\n" \
 SYMBOL_NAME_STR(x) ":\n\t" \
 	"pushl $-1\n\t" \
-        SAVE_ALL \
-        "movl %esp,%eax\n\t" \
-        "pushl %eax\n\t" \
+	SAVE_ALL \
+	"movl %esp,%eax\n\t" \
+	"pushl %eax\n\t" \
 	"call "SYMBOL_NAME_STR(smp_##x)"\n\t" \
-        "addl $4,%esp\n\t" \
+	"addl $4,%esp\n\t" \
 	"jmp ret_from_intr\n");
 
 #endif /* __SMP__ */
 
-#define BUILD_IRQ(chip,nr,mask) \
-asmlinkage void IRQ_NAME(nr); \
-asmlinkage void FAST_IRQ_NAME(nr); \
-asmlinkage void BAD_IRQ_NAME(nr); \
+#define BUILD_COMMON_IRQ() \
 __asm__( \
-"\n"__ALIGN_STR"\n" \
-SYMBOL_NAME_STR(IRQ) #nr "_interrupt:\n\t" \
-	"pushl $-"#nr"-2\n\t" \
+	"\n" __ALIGN_STR"\n" \
+	"common_interrupt:\n\t" \
 	SAVE_ALL \
-	ACK_##chip(mask,(nr&7)) \
-	"movl %esp,%eax\n\t" \
-	"pushl %eax\n\t" \
-	"pushl $" #nr "\n\t" \
-	"call "SYMBOL_NAME_STR(do_IRQ)"\n\t" \
-	"addl $8,%esp\n\t" \
-	UNBLK_##chip(mask) \
-	"jmp ret_from_intr\n" \
-"\n"__ALIGN_STR"\n" \
-SYMBOL_NAME_STR(fast_IRQ) #nr "_interrupt:\n\t" \
-	SAVE_MOST \
-	ACK_##chip(mask,(nr&7)) \
-	"pushl $" #nr "\n\t" \
-	"call "SYMBOL_NAME_STR(do_fast_IRQ)"\n\t" \
-	"addl $4,%esp\n\t" \
-	UNBLK_##chip(mask) \
-	RESTORE_MOST \
-"\n"__ALIGN_STR"\n" \
-SYMBOL_NAME_STR(bad_IRQ) #nr "_interrupt:\n\t" \
-	SAVE_MOST \
-	ACK_##chip(mask,(nr&7)) \
-	RESTORE_MOST);
-	
-#define BUILD_TIMER_IRQ(chip,nr,mask) \
+	"pushl $ret_from_intr\n\t" \
+	"jmp "SYMBOL_NAME_STR(do_IRQ));
+
+#define BUILD_IRQ(chip,nr,mask) \
 asmlinkage void IRQ_NAME(nr); \
-asmlinkage void FAST_IRQ_NAME(nr); \
-asmlinkage void BAD_IRQ_NAME(nr); \
 __asm__( \
 "\n"__ALIGN_STR"\n" \
-SYMBOL_NAME_STR(fast_IRQ) #nr "_interrupt:\n\t" \
-SYMBOL_NAME_STR(bad_IRQ) #nr "_interrupt:\n\t" \
 SYMBOL_NAME_STR(IRQ) #nr "_interrupt:\n\t" \
-	"pushl $-"#nr"-2\n\t" \
-	SAVE_ALL \
-	ACK_##chip(mask,(nr&7)) \
-	"movl %esp,%eax\n\t" \
-	"pushl %eax\n\t" \
-	"pushl $" #nr "\n\t" \
-	"call "SYMBOL_NAME_STR(do_IRQ)"\n\t" \
-	"addl $8,%esp\n\t" \
-	UNBLK_##chip(mask) \
-	"jmp ret_from_intr\n");
+	"pushl $"#nr"-256\n\t" \
+	"jmp common_interrupt");
 
 /*
  * x86 profiling function, SMP safe. We might want to do this in
@@ -276,15 +126,14 @@ static inline void x86_do_profile (unsigned long eip)
 		extern int _stext;
 		eip -= (unsigned long) &_stext;
 		eip >>= prof_shift;
-		if (eip < prof_len)
-			atomic_inc((atomic_t *)&prof_buffer[eip]);
-		else
 		/*
 		 * Dont ignore out-of-bounds EIP values silently,
 		 * put them into the last histogram slot, so if
 		 * present, they will show up as a sharp peak.
 		 */
-			atomic_inc((atomic_t *)&prof_buffer[prof_len-1]);
+		if (eip > prof_len-1)
+			eip = prof_len-1;
+		atomic_inc((atomic_t *)&prof_buffer[eip]);
 	}
 }
 
diff --git a/arch/i386/kernel/process.c b/arch/i386/kernel/process.c
index fe4723951..33842a21f 100644
--- a/arch/i386/kernel/process.c
+++ b/arch/i386/kernel/process.c
@@ -30,6 +30,7 @@
 #include <linux/delay.h>
 #include <linux/smp.h>
 #include <linux/reboot.h>
+#include <linux/init.h>
 #if defined(CONFIG_APM) && defined(CONFIG_APM_POWER_OFF)
 #include <linux/apm_bios.h>
 #endif
@@ -149,7 +150,8 @@ int cpu_idle(void *unused)
 	current->priority = -100;
 	while(1)
 	{
-		if(cpu_data[smp_processor_id()].hlt_works_ok && !hlt_counter && !need_resched)
+		if(cpu_data[smp_processor_id()].hlt_works_ok &&
+		 		!hlt_counter && !need_resched)
 			__asm("hlt");
 		/*
 		 * tq_scheduler currently assumes we're running in a process
@@ -183,7 +185,7 @@ static long no_idt[2] = {0, 0};
 static int reboot_mode = 0;
 static int reboot_thru_bios = 0;
 
-void reboot_setup(char *str, int *ints)
+__initfunc(void reboot_setup(char *str, int *ints))
 {
 	while(1) {
 		switch (*str) {
@@ -324,11 +326,14 @@ void machine_restart(char * __unused)
 
 	pg0 [0] = 7;
 
-	/* Use `swapper_pg_dir' as our page directory.  Don't bother with
-	   `SET_PAGE_DIR' because interrupts are disabled and we're rebooting.
-	   This instruction flushes the TLB. */
+	/*
+	 * Use `swapper_pg_dir' as our page directory.  We bother with
+	 * `SET_PAGE_DIR' because although might be rebooting, but if we change
+	 * the way we set root page dir in the future, then we wont break a
+	 * seldom used feature ;)
+	 */
 
-	__asm__ __volatile__ ("movl %0,%%cr3" : : "a" (swapper_pg_dir) : "memory");
+	SET_PAGE_DIR(current,swapper_pg_dir);
 
 	/* Write 0x1234 to absolute memory location 0x472.  The BIOS reads
 	   this on booting to tell it to "Bypass memory test (also warm
@@ -473,6 +478,8 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
 	int i;
 	struct pt_regs * childregs;
 
+	p->tss.tr = _TSS(nr);
+	p->tss.ldt = _LDT(nr);
 	p->tss.es = KERNEL_DS;
 	p->tss.cs = KERNEL_CS;
 	p->tss.ss = KERNEL_DS;
@@ -480,9 +487,8 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
 	p->tss.fs = USER_DS;
 	p->tss.gs = USER_DS;
 	p->tss.ss0 = KERNEL_DS;
-	p->tss.esp0 = p->kernel_stack_page + PAGE_SIZE;
-	p->tss.tr = _TSS(nr);
-	childregs = ((struct pt_regs *) (p->kernel_stack_page + PAGE_SIZE)) - 1;
+	p->tss.esp0 = 2*PAGE_SIZE + (unsigned long) p;
+	childregs = ((struct pt_regs *) (p->tss.esp0)) - 1;
 	p->tss.esp = (unsigned long) childregs;
 #ifdef __SMP__
 	p->tss.eip = (unsigned long) ret_from_smpfork;
@@ -496,7 +502,6 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
 	childregs->eax = 0;
 	childregs->esp = esp;
 	p->tss.back_link = 0;
-	p->tss.ldt = _LDT(nr);
 	if (p->ldt) {
 		p->ldt = (struct desc_struct*) vmalloc(LDT_ENTRIES*LDT_ENTRY_SIZE);
 		if (p->ldt != NULL)
@@ -512,6 +517,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long esp,
 		p->tss.io_bitmap[i] = ~0;
 	if (last_task_used_math == current)
 		__asm__("clts ; fnsave %0 ; frstor %0":"=m" (p->tss.i387));
+
 	return 0;
 }
 
diff --git a/arch/i386/kernel/ptrace.c b/arch/i386/kernel/ptrace.c
index 920d1bc1c..0dfffd672 100644
--- a/arch/i386/kernel/ptrace.c
+++ b/arch/i386/kernel/ptrace.c
@@ -34,18 +34,6 @@
  */
 #define EFL_OFFSET ((EFL-2)*4-sizeof(struct pt_regs))
 
-/* change a pid into a task struct. */
-static inline struct task_struct * get_task(int pid)
-{
-	int i;
-
-	for (i = 1; i < NR_TASKS; i++) {
-		if (task[i] != NULL && (task[i]->pid == pid))
-			return task[i];
-	}
-	return NULL;
-}
-
 /*
  * this routine will get a word off of the processes privileged stack. 
  * the offset is how far from the base addr as stored in the TSS.  
@@ -95,7 +83,7 @@ static unsigned long get_long(struct task_struct * tsk,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (pgd_none(*pgdir)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -105,7 +93,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -115,7 +103,7 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
@@ -146,7 +134,7 @@ static void put_long(struct task_struct * tsk, struct vm_area_struct * vma, unsi
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (!pgd_present(*pgdir)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -156,7 +144,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -166,12 +154,12 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
 	if (!pte_write(*pgtable)) {
-		do_wp_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 /* this is a hack for non-kernel-mapped video buffers and similar */
@@ -381,7 +369,7 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 	if (pid == 1)		/* you may not mess with init */
 		goto out;
 	ret = -ESRCH;
-	if (!(child = get_task(pid)))
+	if (!(child = find_task_by_pid(pid)))
 		goto out;
 	ret = -EPERM;
 	if (request == PTRACE_ATTACH) {
diff --git a/arch/i386/kernel/setup.c b/arch/i386/kernel/setup.c
index ec5954771..f62744d11 100644
--- a/arch/i386/kernel/setup.c
+++ b/arch/i386/kernel/setup.c
@@ -247,7 +247,7 @@ static const char * i586model(unsigned int nr)
 static const char * i686model(unsigned int nr)
 {
 	static const char *model[] = {
-		"PPro A-step", "Pentium Pro"
+		"PPro A-step", "Pentium Pro", "2", "Pentium II"
 	};
 	if (nr < sizeof(model)/sizeof(char *))
 		return model[nr];
@@ -279,9 +279,10 @@ static const char * getmodel(int x86, int model)
 int get_cpuinfo(char * buffer)
 {
         int i, len = 0;
+	int sep_bug;
         static const char *x86_cap_flags[] = {
                 "fpu", "vme", "de", "pse", "tsc", "msr", "pae", "mce",
-                "cx8", "apic", "10", "11", "mtrr", "pge", "mca", "cmov",
+                "cx8", "apic", "10", "sep", "mtrr", "pge", "mca", "cmov",
                 "16", "17", "18", "19", "20", "21", "22", "mmx",
                 "24", "25", "26", "27", "28", "29", "30", "31"
         };
@@ -321,10 +322,18 @@ int get_cpuinfo(char * buffer)
                         else
                                 len += sprintf(buffer+len, 
                                                "stepping\t: unknown\n");
+
+			sep_bug = CD(have_cpuid) &&
+			          (CD(x86_capability) & 0x800) &&
+			          !memcmp(x86_vendor_id, "GenuineIntel", 12) &&
+			          CD(x86) == 6 &&
+			          CD(x86_model) < 3 &&
+			          CD(x86_mask) < 3;
         
                         len += sprintf(buffer+len,
                                        "fdiv_bug\t: %s\n"
                                        "hlt_bug\t\t: %s\n"
+				       "sep_bug\t\t: %s\n"
                                        "fpu\t\t: %s\n"
                                        "fpu_exception\t: %s\n"
                                        "cpuid\t\t: %s\n"
@@ -332,6 +341,7 @@ int get_cpuinfo(char * buffer)
                                        "flags\t\t:",
                                        CD(fdiv_bug) ? "yes" : "no",
                                        CD(hlt_works_ok) ? "no" : "yes",
+				       sep_bug ? "yes" : "no",
                                        CD(hard_math) ? "yes" : "no",
                                        (CD(hard_math) && ignore_irq13)
                                          ? "yes" : "no",
diff --git a/arch/i386/kernel/signal.c b/arch/i386/kernel/signal.c
index 970c8c5d7..3141c5318 100644
--- a/arch/i386/kernel/signal.c
+++ b/arch/i386/kernel/signal.c
@@ -318,6 +318,14 @@ asmlinkage int do_signal(unsigned long oldmask, struct pt_regs * regs)
 	unsigned long signr;
 	struct sigaction * sa;
 
+	/*
+	 *    We want the common case to go fast, which
+	 * is why we may in certain cases get here from
+	 * kernel mode. Just return without doing anything
+	 * if so.
+	 */
+	if ((regs->xcs & 3) != 3)
+		return 1;
 	mask = ~current->blocked;
 	while ((signr = current->signal & mask)) {
 		/*
@@ -384,10 +392,12 @@ asmlinkage int do_signal(unsigned long oldmask, struct pt_regs * regs)
 
 			case SIGQUIT: case SIGILL: case SIGTRAP:
 			case SIGABRT: case SIGFPE: case SIGSEGV:
+				lock_kernel();
 				if (current->binfmt && current->binfmt->core_dump) {
 					if (current->binfmt->core_dump(signr, regs))
 						signr |= 0x80;
 				}
+				unlock_kernel();
 				/* fall through */
 			default:
 				spin_lock_irq(&current->sigmask_lock);
diff --git a/arch/i386/kernel/smp.c b/arch/i386/kernel/smp.c
index a1590f500..1dc615501 100644
--- a/arch/i386/kernel/smp.c
+++ b/arch/i386/kernel/smp.c
@@ -1,5 +1,5 @@
 /*
- *	Intel MP v1.1/v1.4 specification support routines for multi-pentium 
+ *	Intel MP v1.1/v1.4 specification support routines for multi-pentium
  *	hosts.
  *
  *	(c) 1995 Alan Cox, CymruNET Ltd  <alan@cymru.net>
@@ -46,14 +46,15 @@
 #include <asm/smp.h>
 #include <asm/io.h>
 
+#define __KERNEL_SYSCALLS__
+#include <linux/unistd.h>
+
 #include "irq.h"
 
 extern unsigned long start_kernel, _etext;
 extern void update_one_process( struct task_struct *p,
-                                unsigned long ticks, unsigned long user,
-                                unsigned long system);
-void setup_APIC_clock (void);
-
+				unsigned long ticks, unsigned long user,
+				unsigned long system);
 /*
  *	Some notes on processor bugs:
  *
@@ -67,7 +68,7 @@ void setup_APIC_clock (void);
  *	Pentium
  *		There is a marginal case where REP MOVS on 100MHz SMP
  *	machines with B stepping processors can fail. XXX should provide
- *	an L1cache=Writethrough or L1cache=off option. 
+ *	an L1cache=Writethrough or L1cache=off option.
  *
  *		B stepping CPU's may hang. There are hardware work arounds
  *	for this. We warn about it in case your board doesnt have the work
@@ -91,12 +92,12 @@ void setup_APIC_clock (void);
  *	If this sounds worrying believe me these bugs are ___RARE___ and
  *	there's about nothing of note with C stepping upwards.
  */
- 
- 
+
+
 /*
  *	Why isn't this somewhere standard ??
  */
- 
+
 extern __inline int max(int a,int b)
 {
 	if(a>b)
@@ -121,7 +122,6 @@ struct cpuinfo_x86 cpu_data[NR_CPUS];			/* Per cpu bogomips and other parameters
 static unsigned int num_processors = 1;			/* Internal processor count				*/
 static unsigned long io_apic_addr = 0xFEC00000;		/* Address of the I/O apic (not yet used) 		*/
 unsigned char boot_cpu_id = 0;				/* Processor that is doing the boot up 			*/
-static unsigned char *kstack_base,*kstack_end;		/* Kernel stack list pointers 				*/
 static int smp_activated = 0;				/* Tripped once we need to start cross invalidating 	*/
 int apic_version[NR_CPUS];				/* APIC version number					*/
 static volatile int smp_commenced=0;			/* Tripped when we start scheduling 		    	*/
@@ -129,7 +129,6 @@ unsigned long apic_addr = 0xFEE00000;			/* Address of APIC (defaults to 0xFEE000
 unsigned long nlong = 0;				/* dummy used for apic_reg address + 0x20		*/
 unsigned char *apic_reg=((unsigned char *)(&nlong))-0x20;/* Later set to the ioremap() of the APIC 		*/
 unsigned long apic_retval;				/* Just debugging the assembler.. 			*/
-unsigned char *kernel_stacks[NR_CPUS];			/* Kernel stack pointers for CPU's (debugging)		*/
 
 static volatile unsigned char smp_cpu_in_msg[NR_CPUS];	/* True if this processor is sending an IPI		*/
 
@@ -195,10 +194,10 @@ static inline void ack_APIC_irq (void)
 	apic_write(APIC_EOI, 0);
 }
 
-/* 
+/*
  *	Checksum an MP configuration block.
  */
- 
+
 static int mpf_checksum(unsigned char *mp, int len)
 {
 	int sum=0;
@@ -210,7 +209,7 @@ static int mpf_checksum(unsigned char *mp, int len)
 /*
  *	Processor encoding in an MP configuration block
  */
- 
+
 static char *mpc_family(int family,int model)
 {
 	static char n[32];
@@ -274,11 +273,11 @@ __initfunc(static int smp_read_mpc(struct mp_config_table *mpc))
 
 	/* set the local APIC address */
 	apic_addr = (unsigned long)phys_to_virt((unsigned long)mpc->mpc_lapic);
-	
+
 	/*
 	 *	Now process the configuration blocks.
 	 */
-	 
+	
 	while(count<mpc->mpc_length)
 	{
 		switch(*mpt)
@@ -290,13 +289,13 @@ __initfunc(static int smp_read_mpc(struct mp_config_table *mpc))
 				if(m->mpc_cpuflag&CPU_ENABLED)
 				{
 					printk("Processor #%d %s APIC version %d\n",
-						m->mpc_apicid, 
+						m->mpc_apicid,
 						mpc_family((m->mpc_cpufeature&
 							CPU_FAMILY_MASK)>>8,
 							(m->mpc_cpufeature&
 								CPU_MODEL_MASK)>>4),
 						m->mpc_apicver);
-#ifdef SMP_DEBUG						
+#ifdef SMP_DEBUG
 					if(m->mpc_featureflag&(1<<0))
 						printk("    Floating point unit present.\n");
 					if(m->mpc_featureflag&(1<<7))
@@ -305,7 +304,7 @@ __initfunc(static int smp_read_mpc(struct mp_config_table *mpc))
 						printk("    64 bit compare & exchange supported.\n");
 					if(m->mpc_featureflag&(1<<9))
 						printk("    Internal APIC present.\n");
-#endif						
+#endif
 					if(m->mpc_cpuflag&CPU_BOOTPROCESSOR)
 					{
 						SMP_PRINTK(("    Bootup CPU\n"));
@@ -313,10 +312,10 @@ __initfunc(static int smp_read_mpc(struct mp_config_table *mpc))
 					}
 					else	/* Boot CPU already counted */
 						num_processors++;
-						
+
 					if(m->mpc_apicid>NR_CPUS)
 						printk("Processor #%d unused. (Max %d processors).\n",m->mpc_apicid, NR_CPUS);
-					else						
+					else
 					{
 						cpu_present_map|=(1<<m->mpc_apicid);
 						apic_version[m->mpc_apicid]=m->mpc_apicver;
@@ -337,7 +336,7 @@ __initfunc(static int smp_read_mpc(struct mp_config_table *mpc))
 					str));
 				mpt+=sizeof(*m);
 				count+=sizeof(*m);
-				break; 
+				break;
 			}
 			case MP_IOAPIC:
 			{
@@ -346,20 +345,20 @@ __initfunc(static int smp_read_mpc(struct mp_config_table *mpc))
 				if(m->mpc_flags&MPC_APIC_USABLE)
 				{
 					apics++;
-	                                printk("I/O APIC #%d Version %d at 0x%lX.\n",
-	                                	m->mpc_apicid,m->mpc_apicver,
-	                                	m->mpc_apicaddr);
-	                                io_apic_addr = (unsigned long)phys_to_virt(m->mpc_apicaddr);
-	                        }
-                                mpt+=sizeof(*m);
-                                count+=sizeof(*m); 
-                                break;
+					printk("I/O APIC #%d Version %d at 0x%lX.\n",
+						m->mpc_apicid,m->mpc_apicver,
+						m->mpc_apicaddr);
+					io_apic_addr = (unsigned long)phys_to_virt(m->mpc_apicaddr);
+				}
+				mpt+=sizeof(*m);
+				count+=sizeof(*m);
+				break;
 			}
 			case MP_INTSRC:
 			{
 				struct mpc_config_intsrc *m=
 					(struct mpc_config_intsrc *)mpt;
-				
+
 				mpt+=sizeof(*m);
 				count+=sizeof(*m);
 				break;
@@ -376,29 +375,29 @@ __initfunc(static int smp_read_mpc(struct mp_config_table *mpc))
 	}
 	if(apics>1)
 		printk("Warning: Multiple APIC's not supported.\n");
-	return num_processors;				
+	return num_processors;
 }
 
 /*
  *	Scan the memory blocks for an SMP configuration block.
  */
- 
+
 __initfunc(int smp_scan_config(unsigned long base, unsigned long length))
 {
 	unsigned long *bp=phys_to_virt(base);
 	struct intel_mp_floating *mpf;
-	
+
 	SMP_PRINTK(("Scan SMP from %p for %ld bytes.\n",
 		bp,length));
 	if(sizeof(*mpf)!=16)
 		printk("Error: MPF size\n");
-	
+
 	while(length>0)
 	{
 		if(*bp==SMP_MAGIC_IDENT)
 		{
 			mpf=(struct intel_mp_floating *)bp;
-			if(mpf->mpf_length==1 && 
+			if(mpf->mpf_length==1 &&
 				!mpf_checksum((unsigned char *)bp,16) &&
 				(mpf->mpf_specification == 1
 				 || mpf->mpf_specification == 4) )
@@ -433,7 +432,7 @@ __initfunc(int smp_scan_config(unsigned long base, unsigned long length))
 					 *	We know that page 0 is not
 					 *	used.  Steal it for now!
 					 */
-			 
+			
 					cfg=pg0[0];
 					pg0[0] = (apic_addr | 7);
 					local_flush_tlb();
@@ -451,7 +450,7 @@ __initfunc(int smp_scan_config(unsigned long base, unsigned long length))
  *
  *	END OF HACK   END OF HACK   END OF HACK   END OF HACK   END OF HACK
  *
- */					
+ */
 					/*
 					 *	2 CPUs, numbered 0 & 1.
 					 */
@@ -513,6 +512,7 @@ __initfunc(int smp_scan_config(unsigned long base, unsigned long length))
 				nlong = boot_cpu_id<<24;	/* Dummy 'self' for bootup */
 				cpu_logical_map[0] = boot_cpu_id;
 				global_irq_holder = boot_cpu_id;
+				current->processor = boot_cpu_id;
 
 				printk("Processors: %d\n", num_processors);
 				/*
@@ -534,61 +534,37 @@ __initfunc(int smp_scan_config(unsigned long base, unsigned long length))
 
 extern unsigned char trampoline_data [];
 extern unsigned char trampoline_end  [];
+static unsigned char *trampoline_base;
 
 /*
  *	Currently trivial. Write the real->protected mode
  *	bootstrap into the page concerned. The caller
  *	has made sure it's suitably aligned.
  */
- 
-__initfunc(static void install_trampoline(unsigned char *mp))
+
+__initfunc(static unsigned long setup_trampoline(void))
 {
-	memcpy(mp, trampoline_data, trampoline_end - trampoline_data);
+	memcpy(trampoline_base, trampoline_data, trampoline_end - trampoline_data);
+	return virt_to_phys(trampoline_base);
 }
 
 /*
- *	We are called very early to get the low memory for the trampoline/kernel stacks
- *	This has to be done by mm/init.c to parcel us out nice low memory. We allocate
- *	the kernel stacks at 4K, 8K, 12K... currently (0-03FF is preserved for SMM and
- *	other things).
+ *	We are called very early to get the low memory for the
+ *	SMP bootup trampoline page.
  */
- 
 __initfunc(unsigned long smp_alloc_memory(unsigned long mem_base))
 {
-	int size=(num_processors-1)*PAGE_SIZE;		/* Number of stacks needed */
-
-	/*
-	 *	Our stacks have to be below the 1Mb line, and mem_base on entry
-	 *	is 4K aligned.
-	 */
-	 
-	if(virt_to_phys((void *)(mem_base+size))>=0x9F000)
-		panic("smp_alloc_memory: Insufficient low memory for kernel stacks 0x%lx.\n", mem_base);
-	kstack_base=(void *)mem_base;
-	mem_base+=size;
-	kstack_end=(void *)mem_base;
-	return mem_base;
-}
-	
-/*
- *	Hand out stacks one at a time.
- */
- 
-__initfunc(static void *get_kernel_stack(void))
-{
-	void *stack=kstack_base;
-	if(kstack_base>=kstack_end)
-		return NULL;
-	kstack_base+=PAGE_SIZE;
-	return stack;
+	if (virt_to_phys((void *)mem_base) >= 0x9F000)
+		panic("smp_alloc_memory: Insufficient low memory for kernel trampoline 0x%lx.\n", mem_base);
+	trampoline_base = (void *)mem_base;
+	return mem_base + PAGE_SIZE;
 }
 
-
 /*
  *	The bootstrap kernel entry code has set these up. Save them for
  *	a given CPU
  */
- 
+
 __initfunc(void smp_store_cpu_info(int id))
 {
 	struct cpuinfo_x86 *c=&cpu_data[id];
@@ -615,7 +591,7 @@ __initfunc(void smp_store_cpu_info(int id))
  *	fired off. This allows the BP to have everything in order [we hope].
  *	At the end of this all the AP's will hit the system scheduling and off
  *	we go. Each AP will load the system gdt's and jump through the kernel
- *	init into idle(). At this point the scheduler will one day take over 
+ *	init into idle(). At this point the scheduler will one day take over
  * 	and give them jobs to do. smp_callin is a standard routine
  *	we use to track CPU's as they power up.
  */
@@ -634,74 +610,276 @@ __initfunc(void smp_callin(void))
 	extern void calibrate_delay(void);
 	int cpuid=GET_APIC_ID(apic_read(APIC_ID));
 	unsigned long l;
-	
+
 	/*
 	 *	Activate our APIC
 	 */
-	 
-	SMP_PRINTK(("CALLIN %d\n",smp_processor_id()));
+	
+	SMP_PRINTK(("CALLIN %d %d\n",hard_smp_processor_id(), smp_processor_id()));
  	l=apic_read(APIC_SPIV);
  	l|=(1<<8);		/* Enable */
  	apic_write(APIC_SPIV,l);
 
 	/*
-	 * Set up our APIC timer. 
+	 * Set up our APIC timer.
 	 */
 	setup_APIC_clock ();
 
  	sti();
 	/*
 	 *	Get our bogomips.
-	 */	
+	 */
 	calibrate_delay();
 	SMP_PRINTK(("Stack at about %p\n",&cpuid));
-	
+
 	/*
 	 *	Save our processor parameters
 	 */
  	smp_store_cpu_info(cpuid);
+
 	/*
 	 *	Allow the master to continue.
-	 */	
+	 */
 	set_bit(cpuid, (unsigned long *)&cpu_callin_map[0]);
+}
+
+static int cpucount = 0;
+
+extern int cpu_idle(void * unused);
+
+/*
+ *	Activate a secondary processor.
+ */
+__initfunc(int start_secondary(void *unused))
+{
+	smp_callin();
+	while (!smp_commenced)
+		barrier();
+	return cpu_idle(NULL);
+}
+
+/*
+ * Everything has been set up for the secondary
+ * CPU's - they just need to reload everything
+ * from the task structure
+ */
+__initfunc(void initialize_secondary(void))
+{
+	struct thread_struct * p = &current->tss;
+
 	/*
-	 *	Until we are ready for SMP scheduling
+	 * We don't actually need to load the full TSS,
+	 * basically just the stack pointer and the eip.
 	 */
-	load_ldt(0);
-	local_flush_tlb();
-	
-	while (cpu_number_map[cpuid] == -1)
-		barrier();
+	asm volatile("lldt %%ax": :"a" (p->ldt));
+	asm volatile("ltr %%ax": :"a" (p->tr));
+	asm volatile(
+		"movl %0,%%esp\n\t"
+		"jmp *%1"
+		:
+		:"r" (p->esp),"r" (p->eip));
+}
 
-	while(!task[cpuid] || current_set[cpuid] != task[cpu_number_map[cpuid]])
-		barrier();
+extern struct {
+	void * esp;
+	unsigned short ss;
+} stack_start;
 
-	local_flush_tlb();
-	load_TR(cpu_number_map[cpuid]);
+__initfunc(static void do_boot_cpu(int i))
+{
+	unsigned long cfg;
+	pgd_t maincfg;
+	struct task_struct *idle;
+	unsigned long send_status, accept_status;
+	int timeout, num_starts, j;
+	unsigned long start_eip;
 
-	while(!smp_commenced)
-		barrier();
-	
+	/*
+	 *	We need an idle process for each processor.
+	 */
+
+	kernel_thread(start_secondary, NULL, CLONE_PID);
+	cpucount++;
+
+	idle = task[cpucount];
+	if (!idle)
+		panic("No idle process for CPU %d\n", i);
+
+	idle->processor = i;
+	cpu_logical_map[cpucount] = i;
+	cpu_number_map[i] = cpucount;
+
+	/* start_eip had better be page-aligned! */
+	start_eip = setup_trampoline();
+
+	printk("Booting processor %d eip %lx: ", i, start_eip);	/* So we see what's up   */
+	stack_start.esp = (void *) (1024 + PAGE_SIZE + (char *)idle);
+
+	/*
+	 *	This grunge runs the startup process for
+	 *	the targeted processor.
+	 */
+
+	SMP_PRINTK(("Setting warm reset code and vector.\n"));
+
+	CMOS_WRITE(0xa, 0xf);
 	local_flush_tlb();
+	SMP_PRINTK(("1.\n"));
+	*((volatile unsigned short *) phys_to_virt(0x469)) = start_eip >> 4;
+	SMP_PRINTK(("2.\n"));
+	*((volatile unsigned short *) phys_to_virt(0x467)) = start_eip & 0xf;
+	SMP_PRINTK(("3.\n"));
+
+	maincfg=swapper_pg_dir[0];
+	((unsigned long *)swapper_pg_dir)[0]=0x102007;
+
+	/*
+	 *	Be paranoid about clearing APIC errors.
+	 */
+
+	if ( apic_version[i] & 0xF0 )
+	{
+		apic_write(APIC_ESR, 0);
+		accept_status = (apic_read(APIC_ESR) & 0xEF);
+	}
+
+	/*
+	 *	Status is now clean
+	 */
 	
-	SMP_PRINTK(("Commenced..\n"));
+	send_status = 	0;
+	accept_status = 0;
+
+	/*
+	 *	Starting actual IPI sequence...
+	 */
+
+	SMP_PRINTK(("Asserting INIT.\n"));
+
+	/*
+	 *	Turn INIT on
+	 */
+			
+	cfg=apic_read(APIC_ICR2);
+	cfg&=0x00FFFFFF;
+	apic_write(APIC_ICR2, cfg|SET_APIC_DEST_FIELD(i)); 			/* Target chip     	*/
+	cfg=apic_read(APIC_ICR);
+	cfg&=~0xCDFFF;								/* Clear bits 		*/
+	cfg |= (APIC_DEST_FIELD | APIC_DEST_LEVELTRIG
+		| APIC_DEST_ASSERT | APIC_DEST_DM_INIT);
+	apic_write(APIC_ICR, cfg);						/* Send IPI */
+
+	udelay(200);
+	SMP_PRINTK(("Deasserting INIT.\n"));
+
+	cfg=apic_read(APIC_ICR2);
+	cfg&=0x00FFFFFF;
+	apic_write(APIC_ICR2, cfg|SET_APIC_DEST_FIELD(i));			/* Target chip     	*/
+	cfg=apic_read(APIC_ICR);
+	cfg&=~0xCDFFF;								/* Clear bits 		*/
+	cfg |= (APIC_DEST_FIELD | APIC_DEST_LEVELTRIG
+				| APIC_DEST_DM_INIT);
+	apic_write(APIC_ICR, cfg);						/* Send IPI */
+
+	/*
+	 *	Should we send STARTUP IPIs ?
+	 *
+	 *	Determine this based on the APIC version.
+	 *	If we don't have an integrated APIC, don't
+	 *	send the STARTUP IPIs.
+	 */
+
+	if ( apic_version[i] & 0xF0 )
+		num_starts = 2;
+	else
+		num_starts = 0;
+
+	/*
+	 *	Run STARTUP IPI loop.
+	 */
+
+	for (j = 1; !(send_status || accept_status)
+		    && (j <= num_starts) ; j++)
+	{
+		SMP_PRINTK(("Sending STARTUP #%d.\n",j));
+		apic_write(APIC_ESR, 0);
+		SMP_PRINTK(("After apic_write.\n"));
+
+		/*
+		 *	STARTUP IPI
+		 */
+
+		cfg=apic_read(APIC_ICR2);
+		cfg&=0x00FFFFFF;
+		apic_write(APIC_ICR2, cfg|SET_APIC_DEST_FIELD(i));			/* Target chip     	*/
+		cfg=apic_read(APIC_ICR);
+		cfg&=~0xCDFFF;								/* Clear bits 		*/
+		cfg |= (APIC_DEST_FIELD
+			| APIC_DEST_DM_STARTUP
+			| (start_eip >> 12));						/* Boot on the stack 	*/
+		SMP_PRINTK(("Before start apic_write.\n"));
+		apic_write(APIC_ICR, cfg);						/* Kick the second 	*/
+
+		SMP_PRINTK(("Startup point 1.\n"));
+		timeout = 0;
+		do {
+			SMP_PRINTK(("Sleeping.\n")); udelay(1000000);
+			udelay(10);
+		} while ( (send_status = (apic_read(APIC_ICR) & 0x1000))
+			  && (timeout++ < 1000));
+		udelay(200);
+		accept_status = (apic_read(APIC_ESR) & 0xEF);
+	}
+	SMP_PRINTK(("After Startup.\n"));
+
+	if (send_status)		/* APIC never delivered?? */
+		printk("APIC never delivered???\n");
+	if (accept_status)		/* Send accept error */
+		printk("APIC delivery error (%lx).\n", accept_status);
+
+	if( !(send_status || accept_status) )
+	{
+		for(timeout=0;timeout<50000;timeout++)
+		{
+			if(cpu_callin_map[0]&(1<<i))
+				break;				/* It has booted */
+			udelay(100);				/* Wait 5s total for a response */
+		}
+		if(cpu_callin_map[0]&(1<<i))
+		{
+			/* number CPUs logically, starting from 1 (BSP is 0) */
+#if 0
+			cpu_number_map[i] = cpucount;
+			cpu_logical_map[cpucount] = i;
+#endif
+		}
+		else
+		{
+			if(*((volatile unsigned char *)phys_to_virt(8192))==0xA5)
+				printk("Stuck ??\n");
+			else
+				printk("Not responding.\n");
+		}
+	}
+	SMP_PRINTK(("CPU has booted.\n"));
+
+	swapper_pg_dir[0]=maincfg;
 	local_flush_tlb();
-	sti();
+
+	/* mark "stuck" area as not stuck */
+	*((volatile unsigned long *)phys_to_virt(8192)) = 0;
 }
 
+
 /*
  *	Cycle through the processors sending APIC IPI's to boot each.
  */
- 
+
 __initfunc(void smp_boot_cpus(void))
 {
 	int i;
-	int cpucount=0;
 	unsigned long cfg;
-	pgd_t maincfg;
-	void *stack;
-	extern unsigned long init_user_stack[];
-	
+
 	/*
 	 *	Initialize the logical to physical cpu number mapping
 	 */
@@ -712,12 +890,10 @@ __initfunc(void smp_boot_cpus(void))
 	/*
 	 *	Setup boot CPU information
 	 */
- 
-	kernel_stacks[boot_cpu_id]=(void *)init_user_stack;	/* Set up for boot processor first */
 
 	smp_store_cpu_info(boot_cpu_id);			/* Final full version of the data */
 
-	cpu_present_map |= (1 << smp_processor_id());
+	cpu_present_map |= (1 << hard_smp_processor_id());
 	cpu_number_map[boot_cpu_id] = 0;
 	active_kernel_processor=boot_cpu_id;
 
@@ -744,11 +920,11 @@ __initfunc(void smp_boot_cpus(void))
 	 */
 
 	apic_reg = ioremap(apic_addr,4096);
-	
+
 	if(apic_reg == NULL)
 		panic("Unable to map local apic.\n");
-		
-#ifdef SMP_DEBUG		
+
+#ifdef SMP_DEBUG
 	{
 		int reg;
 
@@ -785,11 +961,11 @@ __initfunc(void smp_boot_cpus(void))
 		SMP_PRINTK(("Getting LVT1: %x\n", reg));
 	}
 #endif
-	
+
 	/*
 	 *	Enable the local APIC
 	 */
- 
+
 	cfg=apic_read(APIC_SPIV);
 	cfg|=(1<<8);		/* Enable APIC */
 	apic_write(APIC_SPIV,cfg);
@@ -798,15 +974,15 @@ __initfunc(void smp_boot_cpus(void))
 
 	/*
 	 * Set up our local APIC timer:
-	 */			
+	 */
 	setup_APIC_clock ();
 
 	/*
 	 *	Now scan the cpu present map and fire up the other CPUs.
 	 */
- 
+
 	SMP_PRINTK(("CPU map: %lx\n", cpu_present_map));
-		
+
 	for(i=0;i<NR_CPUS;i++)
 	{
 		/*
@@ -814,213 +990,17 @@ __initfunc(void smp_boot_cpus(void))
 		 */
 		if (i == boot_cpu_id)
 			continue;
-		
+
 		if ((cpu_present_map & (1 << i))
 		    && (max_cpus < 0 || max_cpus > cpucount+1))
 		{
-			unsigned long send_status, accept_status;
-			int timeout, num_starts, j;
-			
-			/*
-			 *	We need a kernel stack for each processor.
-			 */
-			
-			stack=get_kernel_stack();	/* We allocated these earlier */
-			if(stack==NULL)
-				panic("No memory for processor stacks.\n");
-				
-			kernel_stacks[i]=(void *)phys_to_virt((unsigned long)stack);
-			install_trampoline(stack);
-
-			printk("Booting processor %d stack %p: ",i,stack);			/* So we set what's up   */
-
-			/*				
-			 *	This grunge runs the startup process for
-			 *	the targeted processor.
-			 */
-
-			SMP_PRINTK(("Setting warm reset code and vector.\n"));
-
-			/*
-			 *	Install a writable page 0 entry.
-			 */
-			 
-			cfg=pg0[0];
-			
-			CMOS_WRITE(0xa, 0xf);
-			pg0[0]=7;
-			local_flush_tlb();
-			SMP_PRINTK(("1.\n"));
-			*((volatile unsigned short *) phys_to_virt(0x469)) = ((unsigned long)stack)>>4;
-			SMP_PRINTK(("2.\n"));
-			*((volatile unsigned short *) phys_to_virt(0x467)) = 0;
-			SMP_PRINTK(("3.\n"));
-			
-			/*
-			 *	Protect it again
-			 */
-			 
-			pg0[0]= cfg;
-			local_flush_tlb();
-
-			/*	walken modif
-			 *	enable mapping of the first 4M at virtual
-			 *	address zero
-			 */
-
-			maincfg=swapper_pg_dir[0];
-			((unsigned long *)swapper_pg_dir)[0]=0x102007;
-
-			/* no need to local_flush_tlb :
-			   we are setting this up for the slave processor ! */
-
-			/*
-			 *	Be paranoid about clearing APIC errors.
-			 */
-
-			if ( apic_version[i] & 0xF0 )
-			{
-				apic_write(APIC_ESR, 0);
-				accept_status = (apic_read(APIC_ESR) & 0xEF);
-			}
-			
-			/*
-			 *	Status is now clean
-			 */
-			 
-			send_status = 	0;
-			accept_status = 0;
-
-			/*
-			 *	Starting actual IPI sequence...
-			 */
-
-			SMP_PRINTK(("Asserting INIT.\n"));
-
-			/*
-			 *	Turn INIT on
-			 */
-			 
-			cfg=apic_read(APIC_ICR2);
-			cfg&=0x00FFFFFF;
-			apic_write(APIC_ICR2, cfg|SET_APIC_DEST_FIELD(i)); 			/* Target chip     	*/
-			cfg=apic_read(APIC_ICR);
-			cfg&=~0xCDFFF;								/* Clear bits 		*/
-			cfg |= (APIC_DEST_FIELD | APIC_DEST_LEVELTRIG
-				| APIC_DEST_ASSERT | APIC_DEST_DM_INIT);
-			apic_write(APIC_ICR, cfg);						/* Send IPI */
-
-			udelay(200);
-			SMP_PRINTK(("Deasserting INIT.\n"));
-
-			cfg=apic_read(APIC_ICR2);
-			cfg&=0x00FFFFFF;
-			apic_write(APIC_ICR2, cfg|SET_APIC_DEST_FIELD(i));			/* Target chip     	*/
-			cfg=apic_read(APIC_ICR);
-			cfg&=~0xCDFFF;								/* Clear bits 		*/
-			cfg |= (APIC_DEST_FIELD | APIC_DEST_LEVELTRIG
-				| APIC_DEST_DM_INIT);
-			apic_write(APIC_ICR, cfg);						/* Send IPI */
-			
-			/*
-			 *	Should we send STARTUP IPIs ?
-			 *
-			 *	Determine this based on the APIC version.
-			 *	If we don't have an integrated APIC, don't
-			 *	send the STARTUP IPIs.
-			 */
-
-			if ( apic_version[i] & 0xF0 )
-				num_starts = 2;
-			else
-				num_starts = 0;
-
-			/*
-			 *	Run STARTUP IPI loop.
-			 */
-
-			for (j = 1; !(send_status || accept_status)
-				    && (j <= num_starts) ; j++)
-			{
-				SMP_PRINTK(("Sending STARTUP #%d.\n",j));
-
-				apic_write(APIC_ESR, 0);
-				SMP_PRINTK(("After apic_write.\n"));
-
-				/*
-				 *	STARTUP IPI
-				 */
-
-				cfg=apic_read(APIC_ICR2);
-				cfg&=0x00FFFFFF;
-				apic_write(APIC_ICR2, cfg|SET_APIC_DEST_FIELD(i));			/* Target chip     	*/
-				cfg=apic_read(APIC_ICR);
-				cfg&=~0xCDFFF;								/* Clear bits 		*/
-				cfg |= (APIC_DEST_FIELD
-					| APIC_DEST_DM_STARTUP
-					| (((int)virt_to_phys(stack)) >> 12));					/* Boot on the stack 	*/		
-				SMP_PRINTK(("Before start apic_write.\n"));
-				apic_write(APIC_ICR, cfg);						/* Kick the second 	*/
-
-				SMP_PRINTK(("Startup point 1.\n"));
-				timeout = 0;
-				do {
-				        SMP_PRINTK(("Sleeping.\n")); udelay(1000000);	
-					udelay(10);
-				} while ( (send_status = (apic_read(APIC_ICR) & 0x1000))
-					  && (timeout++ < 1000));
-				udelay(200);
-
-				accept_status = (apic_read(APIC_ESR) & 0xEF);
-			}
-			SMP_PRINTK(("After Startup.\n"));
-
-			if (send_status)		/* APIC never delivered?? */
-				printk("APIC never delivered???\n");
-			if (accept_status)		/* Send accept error */
-				printk("APIC delivery error (%lx).\n", accept_status);
-			
-			if( !(send_status || accept_status) )
-			{
-				for(timeout=0;timeout<50000;timeout++)
-				{
-					if(cpu_callin_map[0]&(1<<i))
-						break;				/* It has booted */
-					udelay(100);				/* Wait 5s total for a response */
-				}
-				if(cpu_callin_map[0]&(1<<i))
-				{
-					cpucount++;
-					/* number CPUs logically, starting from 1 (BSP is 0) */
-					cpu_number_map[i] = cpucount;
-					cpu_logical_map[cpucount] = i;
-				}
-				else
-				{
-					if(*((volatile unsigned char *)phys_to_virt(8192))==0xA5)
-						printk("Stuck ??\n");
-					else
-						printk("Not responding.\n");
-				}
-			}
-			SMP_PRINTK(("CPU has booted.\n"));
-
-                        /*      walken modif
-                         *      restore mapping of the first 4M
-                         */
-
-                        swapper_pg_dir[0]=maincfg;
-                        
-                        local_flush_tlb();
-
-			/* mark "stuck" area as not stuck */
-			*((volatile unsigned long *)phys_to_virt(8192)) = 0;
+			do_boot_cpu(i);
 		}
-		
-		/* 
+
+		/*
 		 *	Make sure we unmap all failed CPUs
 		 */
-		 
+		
 		if (cpu_number_map[i] == -1)
 			cpu_present_map &= ~(1 << i);
 	}
@@ -1056,12 +1036,12 @@ __initfunc(void smp_boot_cpus(void))
 	/*
 	 *	Allow the user to impress friends.
 	 */
-	
+
 	SMP_PRINTK(("Before bogomips.\n"));
 	if(cpucount==0)
 	{
 		printk("Error: only one processor found.\n");
-		cpu_present_map=(1<<smp_processor_id());
+		cpu_present_map=(1<<hard_smp_processor_id());
 	}
 	else
 	{
@@ -1071,8 +1051,8 @@ __initfunc(void smp_boot_cpus(void))
 			if(cpu_present_map&(1<<i))
 				bogosum+=cpu_data[i].udelay_val;
 		}
-		printk("Total of %d processors activated (%lu.%02lu BogoMIPS).\n", 
-			cpucount+1, 
+		printk("Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
+			cpucount+1,
 			(bogosum+2500)/500000,
 			((bogosum+2500)/5000)%100);
 		SMP_PRINTK(("Before bogocount - setting activated=1.\n"));
@@ -1096,7 +1076,7 @@ __initfunc(void smp_boot_cpus(void))
  *	IDE disk problems), and other messages sent with IRQ's enabled in a civilised fashion. That
  *	will also boost performance.
  */
- 
+
 void smp_message_pass(int target, int msg, unsigned long data, int wait)
 {
 	unsigned long flags;
@@ -1109,11 +1089,11 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
 	/*
 	 *	During boot up send no messages
 	 */
-	 
+	
 	if(!smp_activated || !smp_commenced)
 		return;
-		
-	
+
+
 	/*
 	 *	Skip the reschedule if we are waiting to clear a
 	 *	message at this time. The reschedule cannot wait
@@ -1121,7 +1101,7 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
 	 */
 
 	switch (msg) {
-		case MSG_RESCHEDULE:	
+		case MSG_RESCHEDULE:
 			irq = 0x30;
 			if (smp_cpu_in_msg[p])
 				return;
@@ -1148,21 +1128,21 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
 	 *	no data and can occur during a flush.. guess what panic
 	 *	I got to notice this bug...
 	 */
-	 
+	
 	/*
 	 *	We are busy
 	 */
-	 	
-	smp_cpu_in_msg[p]++;
 	
+	smp_cpu_in_msg[p]++;
+
 /*	printk("SMP message pass #%d to %d of %d\n",
 		p, msg, target);*/
-	
+
 	/*
 	 *	Wait for the APIC to become ready - this should never occur. Its
 	 *	a debugging check really.
 	 */
-	 
+	
 	while(ct<1000)
 	{
 		cfg=apic_read(APIC_ICR);
@@ -1171,14 +1151,14 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
 		ct++;
 		udelay(10);
 	}
-	
+
 	/*
 	 *	Just pray... there is nothing more we can do
 	 */
-	 
+	
 	if(ct==1000)
 		printk("CPU #%d: previous IPI still not cleared after 10mS\n", p);
-		
+
 	/*
 	 *	Program the APIC to deliver the IPI
 	 */
@@ -1190,12 +1170,12 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
 	apic_write(APIC_ICR2, cfg|SET_APIC_DEST_FIELD(target));			/* Target chip     		*/
 	cfg=apic_read(APIC_ICR);
 	cfg&=~0xFDFFF;								/* Clear bits 			*/
-	cfg|=APIC_DEST_FIELD|APIC_DEST_DM_FIXED|irq;				/* Send an IRQ 13		*/		
+	cfg|=APIC_DEST_FIELD|APIC_DEST_DM_FIXED|irq;				/* Send an IRQ 13		*/
 
 	/*
 	 *	Set the target requirement
 	 */
-	 
+	
 	if(target==MSG_ALL_BUT_SELF)
 	{
 		cfg|=APIC_DEST_ALLBUT;
@@ -1213,18 +1193,18 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
 		target_map=(1<<target);
 		cpu_callin_map[0]=0;
 	}
-		
+
 	/*
 	 *	Send the IPI. The write to APIC_ICR fires this off.
 	 */
-	 
+	
 	apic_write(APIC_ICR, cfg);
 	__restore_flags(flags);
-	
+
 	/*
 	 *	Spin waiting for completion
 	 */
-	 
+	
 	switch(wait)
 	{
 		int stuck;
@@ -1247,17 +1227,17 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
 					clear_bit(p, &smp_invalidate_needed);
 				--stuck;
 				if (!stuck) {
-					printk("stuck on smp_invalidate_needed IPI wait\n");
+					printk("stuck on smp_invalidate_needed IPI wait (CPU#%d)\n",p);
 					break;
 				}
 			}
 			break;
 	}
-	
+
 	/*
 	 *	Record our completion
 	 */
-	 
+	
 	smp_cpu_in_msg[p]--;
 }
 
@@ -1266,14 +1246,17 @@ void smp_message_pass(int target, int msg, unsigned long data, int wait)
  *	even with IRQ's off. We have to avoid a pair of crossing flushes
  *	or we are doomed.  See the notes about smp_message_pass.
  */
- 
+
 void smp_flush_tlb(void)
 {
 	unsigned long flags;
+
+#if 0
 	if(smp_activated && smp_processor_id()!=active_kernel_processor) {
 		printk("CPU #%d:Attempted flush tlb IPI when not AKP(=%d)\n",smp_processor_id(),active_kernel_processor);
 		*(char *)0=0;
 	}
+#endif
 /*	printk("SMI-");*/
 
 	/*
@@ -1282,30 +1265,30 @@ void smp_flush_tlb(void)
 	 *	may issue a tlb flush. If you break any one of those three change this to an atomic
 	 *	bus locked or.
 	 */
-	
+
 	smp_invalidate_needed=cpu_present_map;
-	
+
 	/*
 	 *	Processors spinning on the lock will see this IRQ late. The smp_invalidate_needed map will
 	 *	ensure they don't do a spurious flush tlb or miss one.
 	 */
-	 
+	
 	__save_flags(flags);
 	__cli();
 	smp_message_pass(MSG_ALL_BUT_SELF, MSG_INVALIDATE_TLB, 0L, 2);
-	
+
 	/*
 	 *	Flush the local TLB
 	 */
-	 
-	local_flush_tlb();
 	
+	local_flush_tlb();
+
 	__restore_flags(flags);
-	
+
 	/*
 	 *	Completed.
 	 */
-	 
+	
 /*	printk("SMID\n");*/
 }
 
@@ -1315,14 +1298,14 @@ void smp_flush_tlb(void)
  *
  * We do profiling in every local tick, statistics/rescheduling
  * happen only every 'profiling multiplier' ticks. The default
- * multiplier is 1 and it can be changed by writing a 4 bytes multiplier
+ * multiplier is 1 and it can be changed by writing the new multiplier
  * value into /proc/profile.
  */
 
 unsigned int prof_multiplier[NR_CPUS];
 unsigned int prof_counter[NR_CPUS];
 
-static inline void smp_local_timer_interrupt(struct pt_regs * regs)
+void smp_local_timer_interrupt(struct pt_regs * regs)
 {
 	int cpu = smp_processor_id();
 
@@ -1367,7 +1350,7 @@ static inline void smp_local_timer_interrupt(struct pt_regs * regs)
 				kstat.cpu_user += user;
 
 			kstat.cpu_system += system;
-		
+
 		} else {
 #ifdef __SMP_PROF__
 			if (test_bit(cpu,&smp_idle_map))
@@ -1386,14 +1369,11 @@ static inline void smp_local_timer_interrupt(struct pt_regs * regs)
 	 * We take the 'long' return path, and there every subsystem
 	 * grabs the apropriate locks (kernel lock/ irq lock).
 	 *
-	 * FIXME: we want to decouple profiling from the 'long path'.
+	 * we might want to decouple profiling from the 'long path',
+	 * and do the profiling totally in assembly.
 	 *
 	 * Currently this isnt too much of an issue (performancewise),
 	 * we can take more than 100K local irqs per second on a 100 MHz P5.
-	 * [ although we notice need_resched too early, thus the way we
-	 *   schedule (deliver signals and handle bhs) changes. ]
-	 *
-	 * Possibly we could solve these problems with 'smart irqs'.
 	 */
 }
 
@@ -1401,6 +1381,9 @@ static inline void smp_local_timer_interrupt(struct pt_regs * regs)
  * Local APIC timer interrupt. This is the most natural way for doing
  * local interrupts, but local timer interrupts can be emulated by
  * broadcast interrupts too. [in case the hw doesnt support APIC timers]
+ *
+ * [ if a single-CPU system runs an SMP kernel then we call the local
+ *   interrupt as well. Thus we cannot inline the local irq ... ]
  */
 void smp_apic_timer_interrupt(struct pt_regs * regs)
 {
@@ -1415,7 +1398,7 @@ void smp_apic_timer_interrupt(struct pt_regs * regs)
 	smp_local_timer_interrupt(regs);
 }
 
-/*	
+/*
  *	Reschedule call back
  */
 asmlinkage void smp_reschedule_interrupt(void)
@@ -1437,11 +1420,11 @@ asmlinkage void smp_reschedule_interrupt(void)
  */
 asmlinkage void smp_invalidate_interrupt(void)
 {
-	if (clear_bit(smp_processor_id(), &smp_invalidate_needed))
+	if (test_and_clear_bit(smp_processor_id(), &smp_invalidate_needed))
 		local_flush_tlb();
 
 	ack_APIC_irq ();
-}	
+}
 
 /*
  *	CPU halt call-back
@@ -1471,11 +1454,11 @@ asmlinkage void smp_stop_cpu_interrupt(void)
  * but we do not accept timer interrupts yet. We only allow the BP
  * to calibrate.
  */
-static unsigned int get_8254_timer_count (void)
+__initfunc(static unsigned int get_8254_timer_count (void))
 {
 	unsigned int count;
 
-        outb_p(0x00, 0x43);
+	outb_p(0x00, 0x43);
 	count = inb_p(0x40);
 	count |= inb_p(0x40) << 8;
 
@@ -1500,7 +1483,7 @@ static unsigned int get_8254_timer_count (void)
 
 void setup_APIC_timer (unsigned int clocks)
 {
-	unsigned long lvtt1_value; 
+	unsigned long lvtt1_value;
 	unsigned int tmp_value;
 
 	/*
@@ -1508,8 +1491,8 @@ void setup_APIC_timer (unsigned int clocks)
 	 * mode. With the IO APIC we can re-route the external timer
 	 * interrupt and broadcast it as an NMI to all CPUs, so no pain.
 	 *
-	 * NOTE: this trap vector (0x41) and the gate in BUILD_SMP_TIMER_INTERRUPT
-	 * should be the same ;)
+	 * NOTE: this trap vector (0x41) and the gate in
+	 * BUILD_SMP_TIMER_INTERRUPT should be the same ;)
 	 */
 	tmp_value = apic_read(APIC_LVTT);
 	lvtt1_value = APIC_LVT_TIMER_PERIODIC | 0x41;
@@ -1526,7 +1509,7 @@ void setup_APIC_timer (unsigned int clocks)
 	apic_write(APIC_TMICT, clocks/APIC_DIVISOR);
 }
 
-void wait_8254_wraparound (void)
+__initfunc(void wait_8254_wraparound (void))
 {
 	unsigned int curr_count, prev_count=~0;
 	int delta;
@@ -1560,11 +1543,12 @@ void wait_8254_wraparound (void)
  * APIC irq that way.
  */
 
-int calibrate_APIC_clock (void)
+__initfunc(int calibrate_APIC_clock (void))
 {
 	unsigned long long t1,t2;
 	long tt1,tt2;
 	long calibration_result;
+	int i;
 
 	printk("calibrating APIC timer ... ");
 
@@ -1589,10 +1573,12 @@ int calibrate_APIC_clock (void)
 	RTDSC(t1);
 	tt1=apic_read(APIC_TMCCT);
 
+#define LOOPS (HZ/10)
 	/*
-	 * lets wait until we get to the next wrapround:
+	 * lets wait LOOPS wraprounds:
 	 */
-	wait_8254_wraparound ();
+	for (i=0; i<LOOPS; i++)
+		wait_8254_wraparound ();
 
 	tt2=apic_read(APIC_TMCCT);
 	RTDSC(t2);
@@ -1605,46 +1591,47 @@ int calibrate_APIC_clock (void)
 	 * underflown to be exact, as the timer counts down ;)
 	 */
 
-	calibration_result = (tt1-tt2)*APIC_DIVISOR;
+	calibration_result = (tt1-tt2)*APIC_DIVISOR/LOOPS;
 
-	printk("\n..... %ld CPU clocks in 1 timer chip tick.\n",
-			 (unsigned long)(t2-t1));
+	SMP_PRINTK(("\n..... %ld CPU clocks in 1 timer chip tick.",
+			 (unsigned long)(t2-t1)/LOOPS));
 
-	printk("..... %ld APIC bus clocks in 1 timer chip tick.\n",
-			 calibration_result);
+	SMP_PRINTK(("\n..... %ld APIC bus clocks in 1 timer chip tick.",
+			 calibration_result));
 
 
-	printk("..... CPU clock speed is %ld.%ld MHz.\n", 
-		((long)(t2-t1))/(1000000/HZ),
-		((long)(t2-t1))%(1000000/HZ)  );
+	printk("\n..... CPU clock speed is %ld.%04ld MHz.\n",
+		((long)(t2-t1)/LOOPS)/(1000000/HZ),
+		((long)(t2-t1)/LOOPS)%(1000000/HZ)  );
 
-	printk("..... APIC bus clock speed is %ld.%ld MHz.\n", 
+	printk("..... APIC bus clock speed is %ld.%04ld MHz.\n",
 		calibration_result/(1000000/HZ),
 		calibration_result%(1000000/HZ)  );
+#undef LOOPS
 
 	return calibration_result;
 }
 
 static unsigned int calibration_result;
 
-void setup_APIC_clock (void)
+__initfunc(void setup_APIC_clock (void))
 {
 	int cpu = smp_processor_id();
-	unsigned long flags; 
+	unsigned long flags;
 
 	static volatile int calibration_lock;
 
 	save_flags(flags);
 	cli();
 
-	printk("setup_APIC_clock() called.\n");
+	SMP_PRINTK(("setup_APIC_clock() called.\n"));
 
 	/*
 	 * [ setup_APIC_clock() is called from all CPUs, but we want
 	 *   to do this part of the setup only once ... and it fits
 	 *   here best ]
 	 */
-	if (!set_bit(0,&calibration_lock)) {
+	if (!test_and_set_bit(0,&calibration_lock)) {
 
 		calibration_result=calibrate_APIC_clock();
 		/*
@@ -1656,9 +1643,9 @@ void setup_APIC_clock (void)
 		/*
 		 * Other CPU is calibrating, wait for finish:
 		 */
-		printk("waiting for other CPU calibrating APIC timer ... ");
+		SMP_PRINTK(("waiting for other CPU calibrating APIC ... "));
 		while (calibration_lock == 1);
-		printk("done, continuing.\n");
+		SMP_PRINTK(("done, continuing.\n"));
 	}
 
 /*
@@ -1669,16 +1656,8 @@ void setup_APIC_clock (void)
 	prof_counter[cpu] = prof_multiplier[cpu] = 1;
 
 	/*
-	 * FIXME: i sporadically see booting problems (keyboard irq is
-	 * lost, looks like the  timer irq isnt working or some irq
-	 * lock is messed up). Once we reboot the bug doesnt showu
-	 * up anymore.
-	 *
-	 * i'm quite certain it's a timing problem/race condition in
-	 * the bootup logic, not a hw bug. It might have been gone
-	 * meanwhile, tell me if you see it.
+	 * We ACK the APIC, just in case there is something pending.
 	 */
-
 	ack_APIC_irq ();
 
 	restore_flags(flags);
@@ -1686,7 +1665,7 @@ void setup_APIC_clock (void)
 
 /*
  * the frequency of the profiling timer can be changed
- * by writing 4 bytes into /proc/profile.
+ * by writing a multiplier value into /proc/profile.
  *
  * usually you want to run this on all CPUs ;)
  */
@@ -1697,8 +1676,8 @@ int setup_profiling_timer (unsigned int multiplier)
 
 	/*
 	 * Sanity check. [at least 500 APIC cycles should be
-	 * between APIC interrupts as a rule of thumb, rather be
-	 * careful as irq flooding renders the system unusable]
+	 * between APIC interrupts as a rule of thumb, to avoid
+	 * irqs flooding us]
 	 */
 	if ( (!multiplier) || (calibration_result/multiplier < 500))
 		return -EINVAL;
diff --git a/arch/i386/kernel/time.c b/arch/i386/kernel/time.c
index 6a7d6b461..e45cc7279 100644
--- a/arch/i386/kernel/time.c
+++ b/arch/i386/kernel/time.c
@@ -23,6 +23,7 @@
 #include <linux/time.h>
 #include <linux/delay.h>
 #include <linux/init.h>
+#include <linux/smp.h>
 
 #include <asm/uaccess.h>
 #include <asm/io.h>
@@ -379,11 +380,15 @@ static inline void timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
 	do_timer(regs);
 /*
  * In the SMP case we use the local APIC timer interrupt to do the
- * profiling.
+ * profiling, except when we simulate SMP mode on a uniprocessor
+ * system, in that case we have to call the local interrupt handler.
  */
 #ifndef __SMP__
 	if (!user_mode(regs))
 		x86_do_profile(regs->eip);
+#else
+	if (!smp_found_config)
+		smp_local_timer_interrupt(regs);
 #endif
 
 	/*
diff --git a/arch/i386/kernel/trampoline.S b/arch/i386/kernel/trampoline.S
index 63bc51c5e..d0a726f6b 100644
--- a/arch/i386/kernel/trampoline.S
+++ b/arch/i386/kernel/trampoline.S
@@ -21,13 +21,9 @@
  *	and IP is zero.  Thus, data addresses need to be absolute
  *	(no relocation) and are taken with regard to r_base.
  *
- *	On the transition to protected mode, this page appears at
- *	address 8192, so protected mode addresses are with regard
- *	to p_base.
- *
  *	If you work on this file, check the object module with objdump
  *	--full-contents --reloc to make sure there are no relocation
- *	entries.
+ *	entries except for the gdt one..
  */
 
 #include <linux/linkage.h>
@@ -39,15 +35,10 @@
 
 ENTRY(trampoline_data)
 r_base = .
-p_base = . - 8192
 
 	mov	%cs, %ax	# Code and data in the same place
 	mov	%ax, %ds
 
-	mov	%ax, %cx	# Pass stack info to the 32bit boot
-	shl	$4, %cx		# Segment -> Offset
-	add	$4096, %cx	# End of page is wanted
-
 	mov	$1, %bx		# Flag an SMP trampoline
 	cli			# We should be safe anyway
 
@@ -71,37 +62,7 @@ idt_48:
 
 gdt_48:
 	.word	0x0800			# gdt limit = 2048, 256 GDT entries
-	.word	gdt - p_base, 0x0	# gdt base = gdt (first SMP CPU)
-					# we load the others with first table
-					# saves rewriting gdt_48 for each
-gdt:
-	.word	0, 0, 0, 0	# dummy
-
-	.word	0, 0, 0, 0	# unused
-
-# walken modif
-
-	.word	0xFFFF		# 4 Gb - (0x100000*0x1000 = 4Gb)
-	.word	0x0000		# base address = 0
-	.word	0x9A00		# code read / exec
-	.word	0x00CF		# granularity = 4096, 386 (+5th nibble of limit)
-
-	.word	0xFFFF		# 4 Gb - (0x100000*0x1000 = 4Gb)
-	.word	0x0000		# base address = 0
-	.word	0x9200		# data read / write
-	.word	0x00CF		# granularity = 4096, 386 (+5th nibble of limit)
-
-# walken modif
-
-#	.word	0x07FF		# 8 Mb - limit = 2047 (2048 * 4096 = 8 Mb)
-#	.word	0x0000		# base address = 0
-#	.word	0x9A00		# code read / exec
-#	.word	0x00C0		# granularity = 4096, 386
-
-#	.word	0x07FF		# 8 Mb - limit = 2047 (2048 * 4096 = 8 Mb)
-#	.word	0x0000		# base address = 0
-#	.word	0x9200		# data read / write
-#	.word	0x00C0		# granularity = 4096, 386
+	.long	gdt-0xc0000000		# gdt base = gdt (first SMP CPU)
 
 .globl SYMBOL_NAME(trampoline_end)
 SYMBOL_NAME_LABEL(trampoline_end)
diff --git a/arch/i386/kernel/traps.c b/arch/i386/kernel/traps.c
index 905cf5b13..696e37004 100644
--- a/arch/i386/kernel/traps.c
+++ b/arch/i386/kernel/traps.c
@@ -23,10 +23,12 @@
 #include <linux/smp.h>
 #include <linux/smp_lock.h>
 #include <linux/init.h>
+#include <linux/delay.h>
 
 #include <asm/system.h>
 #include <asm/uaccess.h>
 #include <asm/io.h>
+#include <asm/spinlock.h>
 
 asmlinkage int system_call(void);
 asmlinkage void lcall7(void);
@@ -121,7 +123,7 @@ static void show_registers(struct pt_regs *regs)
 	unsigned long esp;
 	unsigned short ss;
 	unsigned long *stack, addr, module_start, module_end;
-	extern char start_kernel, _etext;
+	extern char _stext, _etext;
 
 	esp = (unsigned long) &regs->esp;
 	ss = KERNEL_DS;
@@ -129,8 +131,8 @@ static void show_registers(struct pt_regs *regs)
 		esp = regs->esp;
 		ss = regs->xss & 0xffff;
 	}
-	printk("CPU:    %d\n", smp_processor_id());
-	printk("EIP:    %04x:[<%08lx>]\nEFLAGS: %08lx\n", 0xffff & regs->xcs,regs->eip,regs->eflags);
+	printk("CPU:    %d\nEIP:    %04x:[<%08lx>]\nEFLAGS: %08lx\n",
+		smp_processor_id(), 0xffff & regs->xcs, regs->eip, regs->eflags);
 	printk("eax: %08lx   ebx: %08lx   ecx: %08lx   edx: %08lx\n",
 		regs->eax, regs->ebx, regs->ecx, regs->edx);
 	printk("esi: %08lx   edi: %08lx   ebp: %08lx   esp: %08lx\n",
@@ -138,10 +140,8 @@ static void show_registers(struct pt_regs *regs)
 	printk("ds: %04x   es: %04x   ss: %04x\n",
 		regs->xds & 0xffff, regs->xes & 0xffff, ss);
 	store_TR(i);
-	if (STACK_MAGIC != *(unsigned long *)current->kernel_stack_page)
-		printk("Corrupted stack page\n");
 	printk("Process %s (pid: %d, process nr: %d, stackpage=%08lx)\nStack: ",
-		current->comm, current->pid, 0xffff & i, current->kernel_stack_page);
+		current->comm, current->pid, 0xffff & i, 4096+(unsigned long)current);
 	stack = (unsigned long *) esp;
 	for(i=0; i < kstack_depth_to_print; i++) {
 		if (((long) stack & 4095) == 0)
@@ -166,7 +166,7 @@ static void show_registers(struct pt_regs *regs)
 		 * down the cause of the crash will be able to figure
 		 * out the call path that was taken.
 		 */
-		if (((addr >= (unsigned long) &start_kernel) &&
+		if (((addr >= (unsigned long) &_stext) &&
 		     (addr <= (unsigned long) &_etext)) ||
 		    ((addr >= module_start) && (addr <= module_end))) {
 			if (i && ((i % 8) == 0))
@@ -181,13 +181,19 @@ static void show_registers(struct pt_regs *regs)
 	printk("\n");
 }	
 
+spinlock_t die_lock;
+
 /*static*/ void die_if_kernel(const char * str, struct pt_regs * regs, long err)
 {
 	if ((regs->eflags & VM_MASK) || (3 & regs->xcs) == 3)
 		return;
 	console_verbose();
+	spin_lock_irq(&die_lock);
 	printk("%s: %04lx\n", str, err & 0xffff);
 	show_registers(regs);
+do { int i=2000000000; while (i) i--; } while (0);
+do { int i=2000000000; while (i) i--; } while (0);
+	spin_unlock_irq(&die_lock);
 	do_exit(SIGSEGV);
 }
 
@@ -235,18 +241,45 @@ out:
 	unlock_kernel();
 }
 
-asmlinkage void do_nmi(struct pt_regs * regs, long error_code)
+static void mem_parity_error(unsigned char reason, struct pt_regs * regs)
 {
-	printk("NMI\n"); show_registers(regs);
-#ifdef CONFIG_SMP_NMI_INVAL
-	smp_flush_tlb_rcv();
-#else
-#ifndef CONFIG_IGNORE_NMI
 	printk("Uhhuh. NMI received. Dazed and confused, but trying to continue\n");
-	printk("You probably have a hardware problem with your RAM chips or a\n");
-	printk("power saving mode enabled.\n");
-#endif	
-#endif
+	printk("You probably have a hardware problem with your RAM chips\n");
+}	
+
+static void io_check_error(unsigned char reason, struct pt_regs * regs)
+{
+	unsigned long i;
+
+	printk("NMI: IOCK error (debug interrupt?)\n");
+	show_registers(regs);
+
+	/* Re-enable the IOCK line, wait for a few seconds */
+	reason |= 8;
+	outb(reason, 0x61);
+	i = 2000;
+	while (--i) udelay(1000);
+	reason &= ~8;
+	outb(reason, 0x61);
+}
+
+static void unknown_nmi_error(unsigned char reason, struct pt_regs * regs)
+{
+	printk("Uhhuh. NMI received for unknown reason %02x.\n", reason);
+	printk("Dazed and confused, but trying to continue\n");
+	printk("Do you have a strange power saving mode enabled?\n");
+}
+
+asmlinkage void do_nmi(struct pt_regs * regs, long error_code)
+{
+	unsigned char reason = inb(0x61);
+
+	if (reason & 0x80)
+		mem_parity_error(reason, regs);
+	if (reason & 0x40)
+		io_check_error(reason, regs);
+	if (!(reason & 0xc0))
+		unknown_nmi_error(reason, regs);
 }
 
 asmlinkage void do_debug(struct pt_regs * regs, long error_code)
@@ -380,15 +413,7 @@ __initfunc(void trap_init(void))
 {
 	int i;
 	struct desc_struct * p;
-	static int smptrap=0;
-	
-	if(smptrap)
-	{
-		__asm__("pushfl ; andl $0xffffbfff,(%esp) ; popfl");
-		load_ldt(0);
-		return;
-	}
-	smptrap++;
+
 	if (readl(0x0FFFD9) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24))
 		EISA_bus = 1;
 	set_call_gate(&default_ldt,lcall7);
diff --git a/arch/i386/kernel/vm86.c b/arch/i386/kernel/vm86.c
index bfba24327..a09fa6419 100644
--- a/arch/i386/kernel/vm86.c
+++ b/arch/i386/kernel/vm86.c
@@ -81,8 +81,8 @@ asmlinkage struct pt_regs * save_v86_state(struct kernel_vm86_regs * regs)
 		printk("vm86: could not access userspace vm86_info\n");
 		do_exit(SIGSEGV);
 	}
-	current->tss.esp0 = current->saved_kernel_stack;
-	current->saved_kernel_stack = 0;
+	current->tss.esp0 = current->tss.saved_esp0;
+	current->tss.saved_esp0 = 0;
 	ret = KVM86->regs32;
 	unlock_kernel();
 	return ret;
@@ -137,7 +137,7 @@ asmlinkage int sys_vm86old(struct vm86_struct * v86)
 
 	lock_kernel();
 	tsk = current;
-	if (tsk->saved_kernel_stack)
+	if (tsk->tss.saved_esp0)
 		goto out;
 	tmp  = copy_from_user(&info, v86, VM86_REGS_SIZE1);
 	tmp += copy_from_user(&info.regs.VM86_REGS_PART2, &v86->regs.VM86_REGS_PART2,
@@ -187,7 +187,7 @@ asmlinkage int sys_vm86(unsigned long subfunction, struct vm86plus_struct * v86)
 
 	/* we come here only for functions VM86_ENTER, VM86_ENTER_NO_BYPASS */
 	ret = -EPERM;
-	if (tsk->saved_kernel_stack)
+	if (tsk->tss.saved_esp0)
 		goto out;
 	tmp  = copy_from_user(&info, v86, VM86_REGS_SIZE1);
 	tmp += copy_from_user(&info.regs.VM86_REGS_PART2, &v86->regs.VM86_REGS_PART2,
@@ -247,7 +247,7 @@ static void do_sys_vm86(struct kernel_vm86_struct *info, struct task_struct *tsk
  * Save old state, set default return value (%eax) to 0
  */
 	info->regs32->eax = 0;
-	tsk->saved_kernel_stack = tsk->tss.esp0;
+	tsk->tss.saved_esp0 = tsk->tss.esp0;
 	tsk->tss.esp0 = (unsigned long) &info->VM86_TSS_ESP0;
 
 	tsk->tss.screen_bitmap = info->screen_bitmap;
@@ -601,11 +601,17 @@ static inline void free_vm86_irq(int irqnumber)
 static inline int task_valid(struct task_struct *tsk)
 {
 	struct task_struct *p;
+	int ret = 0;
 
+	read_lock(&tasklist_lock);
 	for_each_task(p) {
-		if ((p == tsk) && (p->sig)) return 1;
+		if ((p == tsk) && (p->sig)) {
+			ret = 1;
+			break;
+		}
 	}
-	return 0;
+	read_unlock(&tasklist_lock);
+	return ret;
 }
 
 static inline void handle_irq_zombies(void)
diff --git a/arch/i386/lib/locks.S b/arch/i386/lib/locks.S
index b9d8b0a9c..30a5bc432 100644
--- a/arch/i386/lib/locks.S
+++ b/arch/i386/lib/locks.S
@@ -10,22 +10,25 @@
 	 * %eax contains callers PC and %edx holds this cpu ID.
 	 */
 ENTRY(__lock_kernel)
-	pushl	%eax		! return address
 1:
 	lock
 	btsl	$0, SYMBOL_NAME(kernel_flag)
 	jnc	3f
+	sti
 2:
 	btl	%dl, SYMBOL_NAME(smp_invalidate_needed)
 	jnc	0f
 	lock
 	btrl	%dl, SYMBOL_NAME(smp_invalidate_needed)
 	jnc	0f
+	pushl	%eax
 	movl	%cr3, %eax
 	movl	%eax, %cr3
+	popl	%eax
 0:
 	btl	$0, SYMBOL_NAME(kernel_flag)
 	jc	2b
+	cli
 	jmp	1b
 
 3:
diff --git a/arch/i386/lib/semaphore.S b/arch/i386/lib/semaphore.S
index ff2443c40..faddbf4ec 100644
--- a/arch/i386/lib/semaphore.S
+++ b/arch/i386/lib/semaphore.S
@@ -13,16 +13,17 @@
  * there is contention on the semaphore.
  */
 ENTRY(__down_failed)
-	pushl %eax	/* return address */
+	pushl %eax	/* save %eax */
 	pushl %edx	/* save %edx */
 	pushl %ecx	/* save %ecx (and argument) */
 	call SYMBOL_NAME(__down)
 	popl %ecx	/* restore %ecx (count on __down not changing it) */
 	popl %edx	/* restore %edx */
+	popl %eax	/* restore %eax */
 	ret
 
+/* Don't save/restore %eax, because that will be our return value */
 ENTRY(__down_failed_interruptible)
-	pushl %eax	/* return address */
 	pushl %edx	/* save %edx */
 	pushl %ecx	/* save %ecx (and argument) */
 	call SYMBOL_NAME(__down_interruptible)
@@ -31,10 +32,11 @@ ENTRY(__down_failed_interruptible)
 	ret
 
 ENTRY(__up_wakeup)
-	pushl %eax	/* return address */
+	pushl %eax	/* save %eax */
 	pushl %edx	/* save %edx */
 	pushl %ecx	/* save %ecx (and argument) */
 	call SYMBOL_NAME(__up)
 	popl %ecx	/* restore %ecx (count on __up not changing it) */
 	popl %edx	/* restore %edx */
+	popl %eax	/* restore %eax */
 	ret
diff --git a/arch/i386/mm/fault.c b/arch/i386/mm/fault.c
index 65a4a67d3..b0404a6a9 100644
--- a/arch/i386/mm/fault.c
+++ b/arch/i386/mm/fault.c
@@ -49,7 +49,7 @@ good_area:
 	start &= PAGE_MASK;
 
 	for (;;) {
-		do_wp_page(current, vma, start, 1);
+		handle_mm_fault(current,vma, start, 1);
 		if (!size)
 			break;
 		size--;
@@ -86,10 +86,6 @@ bad_area:
  */
 asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long error_code)
 {
-	void (*handler)(struct task_struct *,
-			struct vm_area_struct *,
-			unsigned long,
-			int);
 	struct task_struct *tsk = current;
 	struct mm_struct *mm = tsk->mm;
 	struct vm_area_struct * vma;
@@ -128,10 +124,8 @@ asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long error_code)
  */
 good_area:
 	write = 0;
-	handler = do_no_page;
 	switch (error_code & 3) {
 		default:	/* 3: write, present */
-			handler = do_wp_page;
 #ifdef TEST_VERIFY_AREA
 			if (regs->cs == KERNEL_CS)
 				printk("WP fault at %08lx\n", regs->eip);
@@ -148,7 +142,7 @@ good_area:
 			if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
 				goto bad_area;
 	}
-	handler(tsk, vma, address, write);
+	handle_mm_fault(tsk, vma, address, write);
 	up(&mm->mmap_sem);
 	/*
 	 * Did it hit the DOS screen memory VA from vm86 mode?
@@ -169,7 +163,10 @@ bad_area:
 
 	/* Are we prepared to handle this fault?  */
 	if ((fixup = search_exception_table(regs->eip)) != 0) {
-		printk(KERN_DEBUG "Exception at [<%lx>] (%lx)\n", regs->eip, fixup);
+		printk(KERN_DEBUG "%s: Exception at [<%lx>] (%lx)\n",
+			current->comm,
+			regs->eip,
+			fixup);
 		regs->eip = fixup;
 		goto out;
 	}
diff --git a/arch/i386/mm/init.c b/arch/i386/mm/init.c
index 173649338..c8371aa81 100644
--- a/arch/i386/mm/init.c
+++ b/arch/i386/mm/init.c
@@ -139,12 +139,23 @@ extern char __init_begin, __init_end;
 #define write_cr4	".byte 0x0f,0x22,0xe0"
 #endif
 
-#define set_in_cr4(x) \
-__asm__(read_cr4 "\n\t" \
-	"orl %0,%%eax\n\t" \
-	write_cr4 \
-	: : "i" (x) \
-	:"ax");
+/*
+ * Save the cr4 feature set we're using (ie
+ * Pentium 4MB enable and PPro Global page
+ * enable), so that any CPU's that boot up
+ * after us can get the correct flags.
+ */
+unsigned long mmu_cr4_features __initdata = 0;
+
+static inline void set_in_cr4(unsigned long mask)
+{
+	mmu_cr4_features |= mask;
+	__asm__(read_cr4 "\n\t"
+		"orl %0,%%eax\n\t"
+		write_cr4
+		: : "irg" (mask)
+		:"ax");
+}
 
 /*
  * paging_init() sets up the page tables - note that the first 4MB are
diff --git a/arch/i386/vmlinux.lds b/arch/i386/vmlinux.lds
index 6d1195013..7a1fd3d08 100644
--- a/arch/i386/vmlinux.lds
+++ b/arch/i386/vmlinux.lds
@@ -13,6 +13,7 @@ SECTIONS
 	*(.fixup)
 	*(.gnu.warning)
 	} = 0x9090
+  .text.lock : { *(.text.lock) }	/* out-of-line lock text */
   .rodata : { *(.rodata) }
   .kstrtab : { *(.kstrtab) }
 
diff --git a/arch/m68k/Makefile b/arch/m68k/Makefile
index 939a80387..7548a8dc8 100644
--- a/arch/m68k/Makefile
+++ b/arch/m68k/Makefile
@@ -24,19 +24,9 @@ ifneq ($(COMPILE_ARCH),$(ARCH))
 	CROSS_COMPILE = m68k-linux-
 endif
 
-#
-# Set these to indicate how to link it..
-#
-# -zmagic:   
-#
-# LINKFLAGS    = -Ttext 0x100000
-#
-# -qmagic (we need to remove the 32 byte header for bootup purposes)
-#
-
-LINKFLAGS = -Ttext 0x1000
+LINKFLAGS = -T $(TOPDIR)/arch/m68k/vmlinux.lds
 
-CFLAGS := $(CFLAGS) -pipe -fno-strength-reduce
+CFLAGS += -pipe -fno-strength-reduce -ffixed-a2
 
 ifdef CONFIG_OPTIMIZE_040
 CFLAGS := $(CFLAGS) -m68040
diff --git a/arch/m68k/amiga/amifb.c b/arch/m68k/amiga/amifb.c
index c97609611..eb72970d7 100644
--- a/arch/m68k/amiga/amifb.c
+++ b/arch/m68k/amiga/amifb.c
@@ -50,12 +50,14 @@
 #include <linux/delay.h>
 #include <linux/config.h>
 #include <linux/interrupt.h>
+#include <linux/fb.h>
+#include <linux/init.h>
+
 #include <asm/uaccess.h>
 #include <asm/system.h>
 #include <asm/irq.h>
 #include <asm/amigahw.h>
 #include <asm/amigaints.h>
-#include <linux/fb.h>
 
 #define DEBUG
 
@@ -1802,7 +1804,7 @@ static int amiga_fb_set_cursorstate(struct fb_cursorstate *state, int con)
 	 * Initialisation
 	 */
 
-struct fb_info *amiga_fb_init(long *mem_start)
+__initfunc(struct fb_info *amiga_fb_init(long *mem_start))
 {
 	int err, tag, i;
 	u_long chipptr;
diff --git a/arch/m68k/amiga/amiints.c b/arch/m68k/amiga/amiints.c
index b527a8c6c..9df5083c1 100644
--- a/arch/m68k/amiga/amiints.c
+++ b/arch/m68k/amiga/amiints.c
@@ -25,6 +25,7 @@
 #include <linux/kernel.h>
 #include <linux/sched.h>
 #include <linux/kernel_stat.h>
+#include <linux/init.h>
 
 #include <asm/system.h>
 #include <asm/irq.h>
@@ -68,7 +69,7 @@ static void ami_badint(int irq, void *dev_id, struct pt_regs *fp)
  * the amiga IRQ handling routines.
  */
 
-void amiga_init_IRQ(void)
+__initfunc(void amiga_init_IRQ(void))
 {
 	int i;
 
diff --git a/arch/m68k/amiga/amikeyb.c b/arch/m68k/amiga/amikeyb.c
index f2e826023..1058270dd 100644
--- a/arch/m68k/amiga/amikeyb.c
+++ b/arch/m68k/amiga/amikeyb.c
@@ -22,6 +22,7 @@
 #include <linux/kd.h>
 #include <linux/random.h>
 #include <linux/kernel.h>
+#include <linux/init.h>
 
 #include <asm/amigaints.h>
 #include <asm/amigahw.h>
@@ -295,7 +296,7 @@ static void keyboard_interrupt(int irq, void *dummy, struct pt_regs *fp)
 	}
 }
 
-int amiga_keyb_init(void)
+__initfunc(int amiga_keyb_init(void))
 {
     if (!AMIGAHW_PRESENT(AMI_KEYBOARD))
         return -EIO;
diff --git a/arch/m68k/amiga/amisound.c b/arch/m68k/amiga/amisound.c
index b2b2e3055..d5656d170 100644
--- a/arch/m68k/amiga/amisound.c
+++ b/arch/m68k/amiga/amisound.c
@@ -10,6 +10,7 @@
 
 #include <linux/sched.h>
 #include <linux/timer.h>
+#include <linux/init.h>
 
 #include <asm/system.h>
 #include <asm/amigahw.h>
@@ -39,7 +40,7 @@ u_short amiga_audio_period = MAX_PERIOD;
 
 static u_long clock_constant;
 
-static void init_sound(void)
+__initfunc(static void init_sound(void))
 {
 	snd_data = amiga_chip_alloc(sizeof(sine_data));
 	if (!snd_data) {
@@ -84,7 +85,7 @@ void amiga_mksound( unsigned int hz, unsigned int ticks )
 		custom.aud[2].audlc = snd_data;
 		custom.aud[2].audlen = sizeof(sine_data)/2;
 		custom.aud[2].audper = (u_short)period;
-		custom.aud[2].audvol = 64; /* maxvol */
+		custom.aud[2].audvol = 32; /* 50% of maxvol */
 	
 		if (ticks) {
 			sound_timer.expires = jiffies + ticks;
diff --git a/arch/m68k/amiga/chipram.c b/arch/m68k/amiga/chipram.c
index 237662c93..50d5a0de0 100644
--- a/arch/m68k/amiga/chipram.c
+++ b/arch/m68k/amiga/chipram.c
@@ -8,6 +8,7 @@
 
 #include <linux/types.h>
 #include <linux/kernel.h>
+#include <linux/init.h>
 #include <asm/amigahw.h>
 
 struct chip_desc {
@@ -21,19 +22,18 @@ struct chip_desc {
 #define DP(ptr) ((struct chip_desc *)(ptr))
 
 u_long amiga_chip_size;
-static unsigned long chipavail; /*MILAN*/
+static unsigned long chipavail;
 
-/*MILAN*/
 unsigned long amiga_chip_avail( void )
 {
 #ifdef DEBUG
-   printk("chip_avail : %ld bytes\n",chipavail);
+	printk("chip_avail : %ld bytes\n",chipavail);
 #endif
-   return chipavail;
+	return chipavail;
 }
 
 
-void amiga_chip_init (void)
+__initfunc(void amiga_chip_init (void))
 {
   struct chip_desc *dp;
 
diff --git a/arch/m68k/amiga/cia.c b/arch/m68k/amiga/cia.c
index 02b1ad564..408dcbf5a 100644
--- a/arch/m68k/amiga/cia.c
+++ b/arch/m68k/amiga/cia.c
@@ -15,6 +15,7 @@
 #include <linux/sched.h>
 #include <linux/errno.h>
 #include <linux/kernel_stat.h>
+#include <linux/init.h>
 
 #include <asm/irq.h>
 #include <asm/amigahw.h>
@@ -152,7 +153,7 @@ static void cia_handler(int irq, void *dev_id, struct pt_regs *fp)
 	amiga_do_irq_list(base->server_irq, fp, &base->server);
 }
 
-void cia_init_IRQ(struct ciabase *base)
+__initfunc(void cia_init_IRQ(struct ciabase *base))
 {
 	int i;
 
diff --git a/arch/m68k/amiga/config.c b/arch/m68k/amiga/config.c
index 8e6ae531c..08396a5bc 100644
--- a/arch/m68k/amiga/config.c
+++ b/arch/m68k/amiga/config.c
@@ -20,6 +20,7 @@
 #include <linux/kd.h>
 #include <linux/tty.h>
 #include <linux/console.h>
+#include <linux/init.h>
 
 #include <asm/bootinfo.h>
 #include <asm/setup.h>
@@ -148,7 +149,7 @@ int amiga_parse_bootinfo(const struct bi_record *record)
      *  Setup the Amiga configuration info
      */
 
-void config_amiga(void)
+__initfunc(void config_amiga(void))
 {
   /* Fill in some default values, if necessary */
   if (amiga_eclock == 0)
@@ -376,8 +377,8 @@ void config_amiga(void)
 
 static unsigned short jiffy_ticks;
 
-static void amiga_sched_init(void (*timer_routine)(int, void *,
-			     struct pt_regs *))
+__initfunc(static void amiga_sched_init(void (*timer_routine)(int, void *,
+					struct pt_regs *)))
 {
 	jiffy_ticks = (amiga_eclock+HZ/2)/HZ;
 
@@ -796,7 +797,7 @@ void amiga_serial_gets(char *s, int len)
 }
 #endif
 
-static void amiga_debug_init(void)
+__initfunc(static void amiga_debug_init(void))
 {
     if (!strcmp( m68k_debug_device, "ser" )) {
         /* no initialization required (?) */
diff --git a/arch/m68k/amiga/cyberfb.c b/arch/m68k/amiga/cyberfb.c
index d74d52c4f..31e2a4608 100644
--- a/arch/m68k/amiga/cyberfb.c
+++ b/arch/m68k/amiga/cyberfb.c
@@ -29,12 +29,13 @@
 #include <linux/tty.h>
 #include <linux/malloc.h>
 #include <linux/delay.h>
+#include <linux/zorro.h>
+#include <linux/fb.h>
+#include <linux/init.h>
 #include <asm/uaccess.h>
 #include <asm/system.h>
 #include <asm/irq.h>
-#include <linux/zorro.h>
 #include <asm/pgtable.h>
-#include <linux/fb.h>
 #include "s3blit.h"
 
 
@@ -1157,7 +1158,7 @@ void Cyber_video_setup(char *options, int *ints)
     *    Initialization
     */
 
-struct fb_info *Cyber_fb_init(long *mem_start)
+__initfunc(struct fb_info *Cyber_fb_init(long *mem_start))
 {
    int err;
    struct Cyber_fb_par par;
diff --git a/arch/m68k/amiga/retz3fb.c b/arch/m68k/amiga/retz3fb.c
index 4885e48f7..49a0853bc 100644
--- a/arch/m68k/amiga/retz3fb.c
+++ b/arch/m68k/amiga/retz3fb.c
@@ -30,1762 +30,12 @@
 #include <linux/malloc.h>
 #include <linux/delay.h>
 #include <linux/fb.h>
-#include <asm/uaccess.h>
-#include <asm/system.h>
-#include <asm/irq.h>
 #include <linux/zorro.h>
-#include <asm/pgtable.h>
-
-#include "retz3fb.h"
-
-/* #define DEBUG if(1) */
-#define DEBUG if(0)
-
-/*
- * Reserve space for one pattern line.
- *
- * For the time being we only support 4MB boards!
- */
-
-#define PAT_MEM_SIZE 16*3
-#define PAT_MEM_OFF  (4*1024*1024 - PAT_MEM_SIZE)
-
-#define arraysize(x)    (sizeof(x)/sizeof(*(x)))
-
-struct retz3_fb_par {
-	int xres;
-	int yres;
-	int xres_vir;
-	int yres_vir;
-	int xoffset;
-	int yoffset;
-	int bpp;
-
-	struct fb_bitfield red;
-	struct fb_bitfield green;
-	struct fb_bitfield blue;
-	struct fb_bitfield transp;
-
-	int pixclock;
-	int left_margin;	/* time from sync to picture	*/
-	int right_margin;	/* time from picture to sync	*/
-	int upper_margin;	/* time from sync to picture	*/
-	int lower_margin;
-	int hsync_len;	/* length of horizontal sync	*/
-	int vsync_len;	/* length of vertical sync	*/
-	int vmode;
-};
-
-struct display_data {
-	long h_total;		/* Horizontal Total */
-	long h_sstart;		/* Horizontal Sync Start */
-	long h_sstop;		/* Horizontal Sync Stop */
-	long h_bstart;		/* Horizontal Blank Start */
-	long h_bstop;		/* Horizontal Blank Stop */
-	long h_dispend;		/* Horizontal Display End */
-	long v_total;		/* Vertical Total */
-	long v_sstart;		/* Vertical Sync Start */
-	long v_sstop;		/* Vertical Sync Stop */
-	long v_bstart;		/* Vertical Blank Start */
-	long v_bstop;		/* Vertical Blank Stop */
-	long v_dispend;		/* Horizontal Display End */
-};
-
-static struct retz3_fb_par current_par;
-
-static int current_par_valid = 0;
-static int currcon = 0;
-
-static struct display disp[MAX_NR_CONSOLES];
-static struct fb_info fb_info;
-
-static int node;        /* node of the /dev/fb?current file */
-
-
-/*
- *    Switch for Chipset Independency
- */
-
-static struct fb_hwswitch {
-
-   /* Initialisation */
-
-   int (*init)(void);
-
-   /* Display Control */
-
-   int (*encode_fix)(struct fb_fix_screeninfo *fix, struct retz3_fb_par *par);
-   int (*decode_var)(struct fb_var_screeninfo *var, struct retz3_fb_par *par);
-   int (*encode_var)(struct fb_var_screeninfo *var, struct retz3_fb_par *par);
-   int (*getcolreg)(unsigned int regno, unsigned int *red, unsigned
-		    int *green, unsigned int *blue, unsigned int *transp);
-   int (*setcolreg)(unsigned int regno, unsigned int red, unsigned int
-		    green, unsigned int blue, unsigned int transp);
-   void (*blank)(int blank);
-} *fbhw;
-
-
-/*
- *    Frame Buffer Name
- */
-
-static char retz3_fb_name[16] = "RetinaZ3";
-
-
-static int z3_key = 0;
-static unsigned char retz3_color_table [256][4];
-static unsigned long z3_mem;
-static unsigned long z3_fbmem;
-static unsigned long z3_size;
-static volatile unsigned char *z3_regs;
-
-static long *memstart;
-
-
-/*
- *    Predefined Video Mode Names
- */
-
-static char *retz3_fb_modenames[] = {
-
-	/*
-	 *    Autodetect (Default) Video Mode
-	 */
-
-	"default",
-
-	/*
-	 *    Predefined Video Modes
-	 */
-
-	"640x480",		/* RetinaZ3 8 bpp */
-	"800x600",		/* RetinaZ3 8 bpp */
-	"1024x768i",
-	"640x480-16",		/* RetinaZ3 16 bpp */
-	"640x480-24",		/* RetinaZ3 24 bpp */
-	
-	/*
-	 *    Dummy Video Modes
-	 */
-
-	"dummy", "dummy", "dummy", "dummy", "dummy", "dummy", "dummy",
-	"dummy", "dummy", "dummy", "dummy", "dummy", "dummy", "dummy",
-	"dummy", "dummy", "dummy", "dummy", "dummy", "dummy",
-
-	/*
-	 *    User Defined Video Modes
-	 *
-	 *    This doesn't work yet!!
-	 */
-
-	"user0", "user1", "user2", "user3",
-	"user4", "user5", "user6", "user7"
-};
-
-/*
- * A small info on how to convert XFree86 timing values into fb
- * timings - by Frank Neumann:
- *
-An XFree86 mode line consists of the following fields:
- "800x600"     50      800  856  976 1040    600  637  643  666
- < name >     DCF       HR  SH1  SH2  HFL     VR  SV1  SV2  VFL
-
-The fields in the fb_var_screeninfo structure are:
-        unsigned long pixclock;         * pixel clock in ps (pico seconds) *
-        unsigned long left_margin;      * time from sync to picture    *
-        unsigned long right_margin;     * time from picture to sync    *
-        unsigned long upper_margin;     * time from sync to picture    *
-        unsigned long lower_margin;
-        unsigned long hsync_len;        * length of horizontal sync    *
-        unsigned long vsync_len;        * length of vertical sync      *
-
-1) Pixelclock:
-   xfree: in MHz
-   fb: In Picoseconds (ps)
-
-   pixclock = 1000000 / DCF
-
-2) horizontal timings:
-   left_margin = HFL - SH2
-   right_margin = SH1 - HR
-   hsync_len = SH2 - SH1
-
-3) vertical timings:
-   upper_margin = VFL - SV2
-   lower_margin = SV1 - VR
-   vsync_len = SV2 - SV1
-
-Good examples for VESA timings can be found in the XFree86 source tree,
-under "programs/Xserver/hw/xfree86/doc/modeDB.txt".
-*/
-
-/*
- *    Predefined Video Mode Definitions
- */
-
-static struct fb_var_screeninfo retz3_fb_predefined[] = {
-
-   /*
-    *    Autodetect (Default) Video Mode
-    */
-
-   { 0, },
-
-   /*
-    *    Predefined Video Modes
-    */
-
-   /*
-    * NB: it is very important to adjust the pixel-clock to the color-depth.
-    */
-
-   {
-	   640, 480, 640, 480, 0, 0, 8, 0,
-	   {0, 8, 0}, {0, 8, 0}, {0, 8, 0}, {0, 0, 0},
-	   0, 0, -1, -1, FB_ACCEL_RETINAZ3, 38461, 28, 32, 12, 10, 96, 2,
-	   FB_SYNC_HOR_HIGH_ACT | FB_SYNC_VERT_HIGH_ACT, FB_VMODE_NONINTERLACED
-   },
-   /*
-    ModeLine "800x600" 36 800 824 896 1024 600 601 603 625
-             < name > DCF HR  SH1 SH2  HFL VR  SV1 SV2 VFL
-    */
-   {
-	   /* 800 x 600, 8 bpp */
-	   800, 600, 800, 600, 0, 0, 8, 0,
-	   {0, 8, 0}, {0, 8, 0}, {0, 8, 0}, {0, 0, 0},
-	   0, 0, -1, -1, FB_ACCEL_RETINAZ3, 27778, 64, 24, 22, 1, 120, 2,
-	   FB_SYNC_COMP_HIGH_ACT|FB_SYNC_VERT_HIGH_ACT, FB_VMODE_NONINTERLACED
-   },
-   /*
-    ModeLine "1024x768i" 45 1024 1064 1224 1264 768 777 785 817 interlace
-              < name >   DCF HR  SH1  SH2  HFL  VR  SV1 SV2 VFL
-    */
-   {
-	   /* 1024 x 768, 8 bpp, interlaced */
-	   1024, 768, 1024, 768, 0, 0, 8, 0,
-	   {0, 8, 0}, {0, 8, 0}, {0, 8, 0}, {0, 0, 0},
-	   0, 0, -1, -1, FB_ACCEL_RETINAZ3, 22222, 40, 40, 32, 9, 160, 8,
-	   FB_SYNC_COMP_HIGH_ACT|FB_SYNC_VERT_HIGH_ACT, FB_VMODE_INTERLACED
-   },
-   {
-	   640, 480, 640, 480, 0, 0, 16, 0,
-	   {11, 5, 0}, {5, 6, 0}, {0, 5, 0}, {0, 0, 0},
-	   0, 0, -1, -1, FB_ACCEL_RETINAZ3, 38461/2, 28, 32, 12, 10, 96, 2,
-	   FB_SYNC_HOR_HIGH_ACT | FB_SYNC_VERT_HIGH_ACT, FB_VMODE_NONINTERLACED
-   },
-   {
-	   640, 480, 640, 480, 0, 0, 24, 0,
-	   {8, 8, 8}, {8, 8, 8}, {8, 8, 8}, {0, 0, 0}, 
-	   0, 0, -1, -1, FB_ACCEL_RETINAZ3, 38461/3, 28, 32, 12, 10, 96, 2,
-	   FB_SYNC_HOR_HIGH_ACT | FB_SYNC_VERT_HIGH_ACT, FB_VMODE_NONINTERLACED
-   },
-
-   /*
-    *    Dummy Video Modes
-    */
-
-   { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, },
-   { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, },
-   { 0, }, { 0, },
-
-   /*
-    *    User Defined Video Modes
-    */
-
-   { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }, { 0, }
-};
-
-
-#define NUM_TOTAL_MODES    arraysize(retz3_fb_predefined)
-#define NUM_PREDEF_MODES   (5)
-
-
-static int z3fb_inverse = 0;
-static int z3fb_mode = 0;
-
-
-/*
- *    Interface used by the world
- */
-
-int retz3_probe(void);
-void retz3_video_setup(char *options, int *ints);
-
-static int retz3_fb_get_fix(struct fb_fix_screeninfo *fix, int con);
-static int retz3_fb_get_var(struct fb_var_screeninfo *var, int con);
-static int retz3_fb_set_var(struct fb_var_screeninfo *var, int con);
-static int retz3_fb_get_cmap(struct fb_cmap *cmap, int kspc, int con);
-static int retz3_fb_set_cmap(struct fb_cmap *cmap, int kspc, int con);
-static int retz3_fb_pan_display(struct fb_var_screeninfo *var, int con);
-static int retz3_fb_ioctl(struct inode *inode, struct file *file,
-			  unsigned int cmd, unsigned long arg, int con);
-
-
-/*
- *    Interface to the low level console driver
- */
-
-struct fb_info *retz3_fb_init(long *mem_start); /* Through amiga_fb_init() */
-static int z3fb_switch(int con);
-static int z3fb_updatevar(int con);
-static void z3fb_blank(int blank);
-static int z3fb_setcmap(struct fb_cmap *cmap, int con);
-
-
-/*
- *    Accelerated Functions used by the low level console driver
- */
-
-void retz3_bitblt(struct fb_var_screeninfo *scr,
-		  unsigned short curx, unsigned short cury, unsigned
-		  short destx, unsigned short desty, unsigned short
-		  width, unsigned short height, unsigned short cmd,
-		  unsigned short mask);
-void retz3_fill(unsigned short x, unsigned short y, unsigned short
-		width, unsigned short height, unsigned short mode,
-		unsigned short color);
-
-/*
- *   Hardware Specific Routines
- */
-
-static int retz3_init(void);
-static int retz3_encode_fix(struct fb_fix_screeninfo *fix,
-                          struct retz3_fb_par *par);
-static int retz3_decode_var(struct fb_var_screeninfo *var,
-                          struct retz3_fb_par *par);
-static int retz3_encode_var(struct fb_var_screeninfo *var,
-                          struct retz3_fb_par *par);
-static int retz3_getcolreg(unsigned int regno, unsigned int *red,
-			   unsigned int *green, unsigned int *blue,
-			   unsigned int *transp);
-static int retz3_setcolreg(unsigned int regno, unsigned int red,
-			   unsigned int green, unsigned int blue,
-			   unsigned int transp);
-static void retz3_blank(int blank);
-
-
-/*
- *    Internal routines
- */
-
-static void retz3_fb_get_par(struct retz3_fb_par *par);
-static void retz3_fb_set_par(struct retz3_fb_par *par);
-static int do_fb_set_var(struct fb_var_screeninfo *var, int isactive);
-static struct fb_cmap *get_default_colormap(int bpp);
-static int do_fb_get_cmap(struct fb_cmap *cmap, struct fb_var_screeninfo *var,
-                          int kspc);
-static int do_fb_set_cmap(struct fb_cmap *cmap, struct fb_var_screeninfo *var,
-                          int kspc);
-static void do_install_cmap(int con);
-static void memcpy_fs(int fsfromto, void *to, void *from, int len);
-static void copy_cmap(struct fb_cmap *from, struct fb_cmap *to, int fsfromto);
-static int alloc_cmap(struct fb_cmap *cmap, int len, int transp);
-static void retz3_fb_set_disp(int con);
-static int get_video_mode(const char *name);
-
-
-/* -------------------- Hardware specific routines -------------------------- */
-
-static unsigned short find_fq(unsigned int freq)
-{
-	unsigned long f;
-	long tmp;
-	long prev = 0x7fffffff;
-	long n2, n1 = 3;
-	unsigned long m;
-	unsigned short res = 0;
-
-	if (freq <= 31250000)
-		n2 = 3;
-	else if (freq <= 62500000)
-		n2 = 2;
-	else if (freq <= 125000000)
-		n2 = 1;
-	else if (freq <= 250000000)
-		n2 = 0;
-	else
-		return(0);
-
-
-	do {
-		f = freq >> (10 - n2);
-
-		m = (f * n1) / (14318180/1024);
-
-		if (m > 129)
-			break;
-
-		tmp =  (((m * 14318180) >> n2) / n1) - freq;
-		if (tmp < 0)
-			tmp = -tmp;
-
-		if (tmp < prev) {
-			prev = tmp;
-			res = (((n2 << 5) | (n1-2)) << 8) | (m-2);
-		}
-
-	} while ( (++n1) <= 21);
-
-	return res;
-}
-
-
-static int retz3_set_video(struct fb_var_screeninfo *var,
-			   struct retz3_fb_par *par)
-{
-	float freq_f;
-	long freq;
-
-	int xres, hfront, hsync, hback;
-	int yres, vfront, vsync, vback;
-	unsigned char tmp;
-	unsigned short best_freq;
-	struct display_data data;
-
-	short clocksel = 0; /* Apparantly this is always zero */
-
-	int bpp = var->bits_per_pixel;
-
-	/*
-	 * XXX
-	 */
-	if (bpp == 24)
-		return 0;
-
-	if ((bpp != 8) && (bpp != 16) && (bpp != 24))
-		return -EFAULT;
-
-	par->xoffset = 0;
-	par->yoffset = 0;
-
-	xres   = var->xres * bpp / 4;
-	hfront = var->right_margin * bpp / 4;
-	hsync  = var->hsync_len * bpp / 4;
-	hback  = var->left_margin * bpp / 4;
-
-	if (var->vmode & FB_VMODE_DOUBLE)
-	{
-		yres = var->yres * 2;
-		vfront = var->lower_margin * 2;
-		vsync  = var->vsync_len * 2;
-		vback  = var->upper_margin * 2;
-	}
-	else if (var->vmode & FB_VMODE_INTERLACED)
-	{
-		yres   = (var->yres + 1) / 2;
-		vfront = (var->lower_margin + 1) / 2;
-		vsync  = (var->vsync_len + 1) / 2;
-		vback  = (var->upper_margin + 1) / 2;
-	}
-	else
-	{
-		yres   = var->yres; /* -1 ? */
-		vfront = var->lower_margin;
-		vsync  = var->vsync_len;
-		vback  = var->upper_margin;
-	}
-
-	data.h_total	= (hback / 8) + (xres / 8) 
-			+ (hfront / 8) + (hsync / 8) - 1 /* + 1 */;
-	data.h_dispend	= ((xres + bpp - 1)/ 8) - 1;
-	data.h_bstart	= xres / 8 /* + 1 */;
-
-	data.h_bstop	= data.h_total+1 + 2 + 1;
-	data.h_sstart	= (xres / 8) + (hfront / 8) + 1;
-	data.h_sstop	= (xres / 8) + (hfront / 8) + (hsync / 8) + 1;
-
-	data.v_total	= yres + vfront + vsync + vback - 1;
-
-	data.v_dispend	= yres - 1;
-	data.v_bstart	= yres;
-
-	data.v_bstop	= data.v_total;
-	data.v_sstart	= yres + vfront - 1 - 2;
-	data.v_sstop	= yres + vfront + vsync - 1;
-
-#if 0 /* testing */
-
-	printk("HBS: %i\n", data.h_bstart);
-	printk("HSS: %i\n", data.h_sstart);
-	printk("HSE: %i\n", data.h_sstop);
-	printk("HBE: %i\n", data.h_bstop);
-	printk("HT: %i\n", data.h_total);
-
-	printk("hsync: %i\n", hsync);
-	printk("hfront: %i\n", hfront);
-	printk("hback: %i\n", hback);
-
-	printk("VBS: %i\n", data.v_bstart);
-	printk("VSS: %i\n", data.v_sstart);
-	printk("VSE: %i\n", data.v_sstop);
-	printk("VBE: %i\n", data.v_bstop);
-	printk("VT: %i\n", data.v_total);
-
-	printk("vsync: %i\n", vsync);
-	printk("vfront: %i\n", vfront);
-	printk("vback: %i\n", vback);
-#endif
-
-	if (data.v_total >= 1024)
-		printk("MAYDAY: v_total >= 1024; bailing out!\n");
-
-	reg_w(GREG_MISC_OUTPUT_W, 0xe3 | ((clocksel & 3) * 0x04));
-	reg_w(GREG_FEATURE_CONTROL_W, 0x00);
-
-	seq_w(SEQ_RESET, 0x00);
-	seq_w(SEQ_RESET, 0x03);	/* reset sequencer logic */
-
-	/*
-	 * CLOCKING_MODE bits:
-	 * 2: This one is only set for certain text-modes, wonder if
-	 *    it may be for EGA-lines? (it was referred to as CLKDIV2)
-	 * (The CL drivers sets it to 0x21 with the comment:
-	 *  FullBandwidth (video off) and 8/9 dot clock)
-	 */
-	seq_w(SEQ_CLOCKING_MODE, 0x01 | 0x00 /* 0x08 */);
-
-	seq_w(SEQ_MAP_MASK, 0x0f);        /* enable writing to plane 0-3 */
-	seq_w(SEQ_CHAR_MAP_SELECT, 0x00); /* doesn't matter in gfx-mode */
-	seq_w(SEQ_MEMORY_MODE, 0x06); /* CL driver says 0x0e for 256 col mode*/
-	seq_w(SEQ_RESET, 0x01);
-	seq_w(SEQ_RESET, 0x03);
-
-	seq_w(SEQ_EXTENDED_ENABLE, 0x05);
-
-	seq_w(SEQ_CURSOR_CONTROL, 0x00);	/* disable cursor */
-	seq_w(SEQ_PRIM_HOST_OFF_HI, 0x00);
-	seq_w(SEQ_PRIM_HOST_OFF_HI, 0x00);
-	seq_w(SEQ_LINEAR_0, 0x4a);
-	seq_w(SEQ_LINEAR_1, 0x00);
-
-	seq_w(SEQ_SEC_HOST_OFF_HI, 0x00);
-	seq_w(SEQ_SEC_HOST_OFF_LO, 0x00);
-	seq_w(SEQ_EXTENDED_MEM_ENA, 0x3 | 0x4 | 0x10 | 0x40);
-
-	/*
-	 * The lower 4 bits (0-3) are used to set the font-width for
-	 * text-mode - DON'T try to set this for gfx-mode.
-	 */
-	seq_w(SEQ_EXT_CLOCK_MODE, 0x10);
-	seq_w(SEQ_EXT_VIDEO_ADDR, 0x03);
-
-	/*
-	 * Extended Pixel Control:
-	 * bit 0:   text-mode=0, gfx-mode=1 (Graphics Byte ?)
-	 * bit 1: (Packed/Nibble Pixel Format ?)
-	 * bit 4-5: depth, 0=1-8bpp, 1=9-16bpp, 2=17-24bpp
-	 */
-	seq_w(SEQ_EXT_PIXEL_CNTL, 0x01 | (((bpp / 8) - 1) << 4));
-
-	seq_w(SEQ_BUS_WIDTH_FEEDB, 0x04);
-	seq_w(SEQ_COLOR_EXP_WFG, 0x01);
-	seq_w(SEQ_COLOR_EXP_WBG, 0x00);
-	seq_w(SEQ_EXT_RW_CONTROL, 0x00);
-	seq_w(SEQ_MISC_FEATURE_SEL, (0x51 | (clocksel & 8)));
-	seq_w(SEQ_COLOR_KEY_CNTL, 0x40);
-	seq_w(SEQ_COLOR_KEY_MATCH0, 0x00);
-	seq_w(SEQ_COLOR_KEY_MATCH1, 0x00);
-	seq_w(SEQ_COLOR_KEY_MATCH2, 0x00);
-	seq_w(SEQ_CRC_CONTROL, 0x00);
-	seq_w(SEQ_PERF_SELECT, 0x10);
-	seq_w(SEQ_ACM_APERTURE_1, 0x00);
-	seq_w(SEQ_ACM_APERTURE_2, 0x30);
-	seq_w(SEQ_ACM_APERTURE_3, 0x00);
-	seq_w(SEQ_MEMORY_MAP_CNTL, 0x03);
-
-
-	/* unlock register CRT0..CRT7 */
-	crt_w(CRT_END_VER_RETR, (data.v_sstop & 0x0f) | 0x20);
-
-	/* Zuerst zu schreibende Werte nur per printk ausgeben */
-	DEBUG printk("CRT_HOR_TOTAL: %ld\n", data.h_total);
-	crt_w(CRT_HOR_TOTAL, data.h_total & 0xff);
-
-	DEBUG printk("CRT_HOR_DISP_ENA_END: %ld\n", data.h_dispend);
-	crt_w(CRT_HOR_DISP_ENA_END, (data.h_dispend) & 0xff);
-
-	DEBUG printk("CRT_START_HOR_BLANK: %ld\n", data.h_bstart);
-	crt_w(CRT_START_HOR_BLANK, data.h_bstart & 0xff);
-
-	DEBUG printk("CRT_END_HOR_BLANK: 128+%ld\n", data.h_bstop % 32);
-	crt_w(CRT_END_HOR_BLANK,  0x80 | (data.h_bstop & 0x1f));
-
-	DEBUG printk("CRT_START_HOR_RETR: %ld\n", data.h_sstart);
-	crt_w(CRT_START_HOR_RETR, data.h_sstart & 0xff);
-
-	tmp = (data.h_sstop & 0x1f);
-	if (data.h_bstop & 0x20)
-		tmp |= 0x80;
-	DEBUG printk("CRT_END_HOR_RETR: %d\n", tmp);
-	crt_w(CRT_END_HOR_RETR, tmp);
-
-	DEBUG printk("CRT_VER_TOTAL: %ld\n", data.v_total & 0xff);
-	crt_w(CRT_VER_TOTAL, (data.v_total & 0xff));
-
-	tmp = 0x10;  /* LineCompare bit #9 */
-	if (data.v_total & 256)
-		tmp |= 0x01;
-	if (data.v_dispend & 256)
-		tmp |= 0x02;
-	if (data.v_sstart & 256)
-		tmp |= 0x04;
-	if (data.v_bstart & 256)
-		tmp |= 0x08;
-	if (data.v_total & 512)
-		tmp |= 0x20;
-	if (data.v_dispend & 512)
-		tmp |= 0x40;
-	if (data.v_sstart & 512)
-		tmp |= 0x80;
-	DEBUG printk("CRT_OVERFLOW: %d\n", tmp);
-	crt_w(CRT_OVERFLOW, tmp);
-
-	crt_w(CRT_PRESET_ROW_SCAN, 0x00); /* not CL !!! */
-
-	tmp = 0x40; /* LineCompare bit #8 */
-	if (data.v_bstart & 512)
-		tmp |= 0x20;
-	if (var->vmode & FB_VMODE_DOUBLE)
-		tmp |= 0x80;
- 	DEBUG printk("CRT_MAX_SCAN_LINE: %d\n", tmp);
-	crt_w(CRT_MAX_SCAN_LINE, tmp);
-
-	crt_w(CRT_CURSOR_START, 0x00);
-	crt_w(CRT_CURSOR_END, 8 & 0x1f); /* font height */
-
-	crt_w(CRT_START_ADDR_HIGH, 0x00);
-	crt_w(CRT_START_ADDR_LOW, 0x00);
-
-	crt_w(CRT_CURSOR_LOC_HIGH, 0x00);
-	crt_w(CRT_CURSOR_LOC_LOW, 0x00);
-
- 	DEBUG printk("CRT_START_VER_RETR: %ld\n", data.v_sstart & 0xff);
-	crt_w(CRT_START_VER_RETR, (data.v_sstart & 0xff));
-
-#if 1
-	/* 5 refresh cycles per scanline */
-	DEBUG printk("CRT_END_VER_RETR: 64+32+%ld\n", data.v_sstop % 16);
-	crt_w(CRT_END_VER_RETR, ((data.v_sstop & 0x0f) | 0x40 | 0x20));
-#else
-	DEBUG printk("CRT_END_VER_RETR: 128+32+%ld\n", data.v_sstop % 16);
-	crt_w(CRT_END_VER_RETR, ((data.v_sstop & 0x0f) | 128 | 32));
-#endif
-	DEBUG printk("CRT_VER_DISP_ENA_END: %ld\n", data.v_dispend & 0xff);
-	crt_w(CRT_VER_DISP_ENA_END, (data.v_dispend & 0xff));
-
-	DEBUG printk("CRT_START_VER_BLANK: %ld\n", data.v_bstart & 0xff);
-	crt_w(CRT_START_VER_BLANK, (data.v_bstart & 0xff));
-
-	DEBUG printk("CRT_END_VER_BLANK: %ld\n", data.v_bstop & 0xff);
-	crt_w(CRT_END_VER_BLANK, (data.v_bstop & 0xff));
-
-	DEBUG printk("CRT_MODE_CONTROL: 0xe3\n");
-	crt_w(CRT_MODE_CONTROL, 0xe3);
-
-	DEBUG printk("CRT_LINE_COMPARE: 0xff\n");
-	crt_w(CRT_LINE_COMPARE, 0xff);
-
-	tmp = (var->xres_virtual / 8) * (bpp / 8);
-	crt_w(CRT_OFFSET, tmp);
-
-	crt_w(CRT_UNDERLINE_LOC, 0x07); /* probably font-height - 1 */
-
-	tmp = 0x20;			/* Enable extended end bits */
-	if (data.h_total & 0x100)
-		tmp |= 0x01;
-	if ((data.h_dispend) & 0x100)
-		tmp |= 0x02;
-	if (data.h_bstart & 0x100)
-		tmp |= 0x04;
-	if (data.h_sstart & 0x100)
-		tmp |= 0x08;
-	if (var->vmode & FB_VMODE_INTERLACED)
-		tmp |= 0x10;
- 	DEBUG printk("CRT_EXT_HOR_TIMING1: %d\n", tmp);
-	crt_w(CRT_EXT_HOR_TIMING1, tmp);
-
-	tmp = 0x00;
-	if (((var->xres_virtual / 8) * (bpp / 8)) & 0x100)
-		tmp |= 0x10;
-	crt_w(CRT_EXT_START_ADDR, tmp);
-
-	tmp = 0x00;
-	if (data.h_total & 0x200)
-		tmp |= 0x01;
-	if ((data.h_dispend) & 0x200)
-		tmp |= 0x02;
-	if (data.h_bstart & 0x200)
-		tmp |= 0x04;
-	if (data.h_sstart & 0x200)
-		tmp |= 0x08;
-	tmp |= ((data.h_bstop & 0xc0) >> 2);
-	tmp |= ((data.h_sstop & 0x60) << 1);
-	crt_w(CRT_EXT_HOR_TIMING2, tmp);
- 	DEBUG printk("CRT_EXT_HOR_TIMING2: %d\n", tmp);
-
-	tmp = 0x10;			/* Line compare bit 10 */
-	if (data.v_total & 0x400)
-		tmp |= 0x01;
-	if ((data.v_dispend) & 0x400)
-		tmp |= 0x02;
-	if (data.v_bstart & 0x400)
-		tmp |= 0x04;
-	if (data.v_sstart & 0x400)
-		tmp |= 0x08;
-	tmp |= ((data.v_bstop & 0x300) >> 3);
-	if (data.v_sstop & 0x10)
-		tmp |= 0x80;
-	crt_w(CRT_EXT_VER_TIMING, tmp);
- 	DEBUG printk("CRT_EXT_VER_TIMING: %d\n", tmp);
-
-	crt_w(CRT_MONITOR_POWER, 0x00);
-
-	/*
-	 * Convert from ps to Hz.
-	 */
-	freq_f = (1.0/(float)var->pixclock) * 1000000000;
-	freq = ((long)freq_f) * 1000;
-
-	best_freq = find_fq(freq);
-	pll_w(0x02, best_freq);
-	best_freq = find_fq(61000000);
-	pll_w(0x0a, best_freq);
-	pll_w(0x0e, 0x22);
-
-	gfx_w(GFX_SET_RESET, 0x00);
-	gfx_w(GFX_ENABLE_SET_RESET, 0x00);
-	gfx_w(GFX_COLOR_COMPARE, 0x00);
-	gfx_w(GFX_DATA_ROTATE, 0x00);
-	gfx_w(GFX_READ_MAP_SELECT, 0x00);
-	gfx_w(GFX_GRAPHICS_MODE, 0x00);
-	gfx_w(GFX_MISC, 0x05);
-	gfx_w(GFX_COLOR_XCARE, 0x0f);
-	gfx_w(GFX_BITMASK, 0xff);
-
-	reg_r(ACT_ADDRESS_RESET);
-	attr_w(ACT_PALETTE0 , 0x00);
-	attr_w(ACT_PALETTE1 , 0x01);
-	attr_w(ACT_PALETTE2 , 0x02);
-	attr_w(ACT_PALETTE3 , 0x03);
-	attr_w(ACT_PALETTE4 , 0x04);
-	attr_w(ACT_PALETTE5 , 0x05);
-	attr_w(ACT_PALETTE6 , 0x06);
-	attr_w(ACT_PALETTE7 , 0x07);
-	attr_w(ACT_PALETTE8 , 0x08);
-	attr_w(ACT_PALETTE9 , 0x09);
-	attr_w(ACT_PALETTE10, 0x0a);
-	attr_w(ACT_PALETTE11, 0x0b);
-	attr_w(ACT_PALETTE12, 0x0c);
-	attr_w(ACT_PALETTE13, 0x0d);
-	attr_w(ACT_PALETTE14, 0x0e);
-	attr_w(ACT_PALETTE15, 0x0f);
-	reg_r(ACT_ADDRESS_RESET);
-
-	attr_w(ACT_ATTR_MODE_CNTL, 0x09); /* 0x01 for CL */
-
-	attr_w(ACT_OVERSCAN_COLOR, 0x00);
-	attr_w(ACT_COLOR_PLANE_ENA, 0x0f);
-	attr_w(ACT_HOR_PEL_PANNING, 0x00);
-	attr_w(ACT_COLOR_SELECT, 0x00);
-
-	reg_r(ACT_ADDRESS_RESET);
-	reg_w(ACT_DATA, 0x20);
-
-	reg_w(VDAC_MASK, 0xff);
-
-	/*
-	 * Extended palette adressing ???
-	 */
-	switch (bpp){
-	case 8:
-		reg_w(0x83c6, 0x00); 
-		break;
-	case 16:
-		reg_w(0x83c6, 0x60);
-		break;
-	case 24:
-		reg_w(0x83c6, 0xe0);
-		break;
-	default:
-		printk("Illegal color-depth: %i\n", bpp);
-	}
-
-	reg_w(VDAC_ADDRESS, 0x00);
-
-	seq_w(SEQ_MAP_MASK, 0x0f );  
-
-	return 0;
-}
-
-/*
- *    Initialization
- *
- *    Set the default video mode for this chipset. If a video mode was
- *    specified on the command line, it will override the default mode.
- */
-
-static int retz3_init(void)
-{
-	int i;
-#if 0
-	volatile unsigned long *CursorBase;
-#endif
-	unsigned long board_addr, board_size;
-	struct ConfigDev *cd;
-
-	cd = zorro_get_board (z3_key);
-	zorro_config_board (z3_key, 0);
-	board_addr = (unsigned long)cd->cd_BoardAddr;
-	board_size = (unsigned long)cd->cd_BoardSize;
-
-	for (i = 0; i < 256; i++){
-		for (i = 0; i < 256; i++){
-			retz3_color_table [i][0] = i;
-			retz3_color_table [i][1] = i;
-			retz3_color_table [i][2] = i;
-			retz3_color_table [i][3] = 0;
-		}
-	}
-
-	*memstart = (*memstart + PAGE_SIZE - 1) & ~(PAGE_SIZE - 1);
-
-	z3_mem = kernel_map (board_addr, board_size,
-			     KERNELMAP_NOCACHE_SER, memstart);
-
-	z3_regs = (char*) z3_mem;
-	z3_fbmem = z3_mem + VIDEO_MEM_OFFSET;
-
-	/* Get memory size - for now we asume its a 4MB board */
-
-	z3_size = 0x00400000; /* 4 MB */
-
-	memset ((char*)z3_fbmem, 0, z3_size);
-
-	/* Disable hardware cursor */
-
-	seq_w(SEQ_CURSOR_Y_INDEX, 0x00);
-
-
-#if 0
-	/* Initialize hardware cursor */
-	CursorBase = (unsigned long *)((char *)(z3_mem) + z3_size - 0x400);
-	for (i=0; i < 8; i++){
-		*(CursorBase  +(i*4)) = 0xffffff00;
-		*(CursorBase+1+(i*4)) = 0xffff0000;
-		*(CursorBase+2+(i*4)) = 0xffff0000;
-		*(CursorBase+3+(i*4)) = 0xffff0000;
-	}
-	for (i=8; i < 64; i++){
-		*(CursorBase  +(i*4)) = 0xffff0000;
-		*(CursorBase+1+(i*4)) = 0xffff0000;
-		*(CursorBase+2+(i*4)) = 0xffff0000;
-		*(CursorBase+3+(i*4)) = 0xffff0000;
-	}
-#endif
-
-	retz3_setcolreg (255, 56, 100, 160, 0);
-	retz3_setcolreg (254, 0, 0, 0, 0);
-
-	return 0;
-}
-
-
-/*
- *    This function should fill in the `fix' structure based on the
- *    values in the `par' structure.
- */
-
-static int retz3_encode_fix(struct fb_fix_screeninfo *fix,
-                          struct retz3_fb_par *par)
-{
-	int i;
-
-	strcpy(fix->id, retz3_fb_name);
-	fix->smem_start = z3_fbmem;
-	fix->smem_len = z3_size;
-
-	fix->type = FB_TYPE_PACKED_PIXELS;
-	fix->type_aux = 0;
-	if (par->bpp == 8)
-		fix->visual = FB_VISUAL_PSEUDOCOLOR;
-	else
-		fix->visual = FB_VISUAL_DIRECTCOLOR;
-
-	fix->xpanstep = 0;
-	fix->ypanstep = 0;
-	fix->ywrapstep = 0;
-	fix->line_length = 0;
-
-	for (i = 0; i < arraysize(fix->reserved); i++)
-		fix->reserved[i] = 0;
-
-	return 0;
-}
-
-
-/*
- *    Get the video params out of `var'. If a value doesn't fit, round
- *    it up, if it's too big, return -EINVAL.
- */
-
-static int retz3_decode_var(struct fb_var_screeninfo *var,
-			    struct retz3_fb_par *par)
-{
-	par->xres = var->xres;
-	par->yres = var->yres;
-	par->xres_vir = var->xres_virtual;
-	par->yres_vir = var->yres_virtual;
-	par->bpp = var->bits_per_pixel;
-	par->pixclock = var->pixclock;
-	par->vmode = var->vmode;
-
-	par->red = var->red;
-	par->green = var->green;
-	par->blue = var->blue;
-	par->transp = var->transp;
-
-	par->left_margin = var->left_margin;
-	par->right_margin = var->right_margin;
-	par->upper_margin = var->upper_margin;
-	par->lower_margin = var->lower_margin;
-	par->hsync_len = var->hsync_len;
-	par->vsync_len = var->vsync_len;
-
-	return 0;
-}
-
-
-/*
- *    Fill the `var' structure based on the values in `par' and maybe
- *    other values read out of the hardware.
- */
-
-static int retz3_encode_var(struct fb_var_screeninfo *var,
-			    struct retz3_fb_par *par)
-{
-	int i;
-
-	var->xres = par->xres;
-	var->yres = par->yres;
-	var->xres_virtual = par->xres_vir;
-	var->yres_virtual = par->yres_vir;
-	var->xoffset = 0;
-	var->yoffset = 0;
-
-	var->bits_per_pixel = par->bpp;
-	var->grayscale = 0;
-
-	var->red = par->red;
-	var->green = par->green;
-	var->blue = par->blue;
-	var->transp = par->transp;
-
-	var->nonstd = 0;
-	var->activate = 0;
-
-	var->height = -1;
-	var->width = -1;
-
-	var->accel = FB_ACCEL_RETINAZ3;
-
-	var->pixclock = par->pixclock;
-
-	var->sync = 0;				/* ??? */
-	var->left_margin = par->left_margin;
-	var->right_margin = par->right_margin;
-	var->upper_margin = par->upper_margin;
-	var->lower_margin = par->lower_margin;
-	var->hsync_len = par->hsync_len;
-	var->vsync_len = par->vsync_len;
-
-	for (i = 0; i < arraysize(var->reserved); i++)
-		var->reserved[i] = 0;
-
-	var->vmode = par->vmode;
-	return 0;
-}
-
-
-/*
- *    Set a single color register. The values supplied are already
- *    rounded down to the hardware's capabilities (according to the
- *    entries in the var structure). Return != 0 for invalid regno.
- */
-
-static int retz3_setcolreg(unsigned int regno, unsigned int red,
-			   unsigned int green, unsigned int blue,
-			   unsigned int transp)
-{
-	/* We'll get to this */
-
-	if (regno > 255)
-		return 1;
-
-	retz3_color_table [regno][0] = red & 0xff;
-	retz3_color_table [regno][1] = green & 0xff;
-	retz3_color_table [regno][2] = blue & 0xff;
-	retz3_color_table [regno][3] = transp;
-
-	reg_w(VDAC_ADDRESS_W, regno);
-	reg_w(VDAC_DATA, (red & 0xff) >> 2);
-	reg_w(VDAC_DATA, (green & 0xff) >> 2);
-	reg_w(VDAC_DATA, (blue & 0xff) >> 2);
-
-	return 0;
-}
-
-
-/*
- *    Read a single color register and split it into
- *    colors/transparent. Return != 0 for invalid regno.
- */
-
-static int retz3_getcolreg(unsigned int regno, unsigned int *red,
-			   unsigned int *green, unsigned int *blue,
-			   unsigned int *transp)
-{
-	if (regno > 255)
-		return 1;
-	*red    = retz3_color_table [regno][0];
-	*green  = retz3_color_table [regno][1];
-	*blue   = retz3_color_table [regno][2];
-	*transp = retz3_color_table [regno][3];
-	return 0;
-}
-
-
-/*
- *    (Un)Blank the screen
- */
-
-void retz3_blank(int blank)
-{
-	int i;
-
-	if (blank)
-		for (i = 0; i < 256; i++){
-			reg_w(VDAC_ADDRESS_W, i);
-			reg_w(VDAC_DATA, 0);
-			reg_w(VDAC_DATA, 0);
-			reg_w(VDAC_DATA, 0);
-		}
-	else
-		for (i = 0; i < 256; i++){
-			reg_w(VDAC_ADDRESS_W, i);
-			reg_w(VDAC_DATA, retz3_color_table [i][0] >> 2);
-			reg_w(VDAC_DATA, retz3_color_table [i][1] >> 2);
-			reg_w(VDAC_DATA, retz3_color_table [i][2] >> 2);
-		}
-}
-
-
-void retz3_bitblt (struct fb_var_screeninfo *var,
-		   unsigned short srcx, unsigned short srcy, unsigned
-		   short destx, unsigned short desty, unsigned short
-		   width, unsigned short height, unsigned short cmd,
-		   unsigned short mask)
-{
-
-	volatile unsigned long *acm = (unsigned long *) (z3_mem + ACM_OFFSET);
-	unsigned long *pattern = (unsigned long *)(z3_fbmem + PAT_MEM_OFF);
-
-	unsigned short mod;
-	unsigned long tmp;
-	unsigned long pat, src, dst;
-	unsigned char blt_status;
-
-	int i, xres_virtual = var->xres_virtual;
-	short bpp = (var->bits_per_pixel & 0xff);
-
-	if (bpp < 8)
-		bpp = 8;
-
-	tmp = mask | (mask << 16);
-
-#if 0
-	/*
-	 * Check for blitter finished before we start messing with the
-	 * pattern.
-	 */
-	do{
-		blt_status = *(((volatile unsigned char *)acm) +
-			       (ACM_START_STATUS + 2));
-	}while ((blt_status & 1) == 0);
-#endif
-
-	i = 0;
-	do{
-		*pattern++ = tmp;
-	}while(i++ < bpp/4);
-
-	tmp = cmd << 8;
-	*(acm + ACM_RASTEROP_ROTATION/4) = tmp;
-
-	mod = 0xc0c2;
-
-	pat = 8 * PAT_MEM_OFF;
-	dst = bpp * (destx + desty * xres_virtual);
-
-	/*
-	 * Source is not set for clear.
-	 */
-	if ((cmd != Z3BLTclear) && (cmd != Z3BLTset)) {
-		src = bpp * (srcx + srcy * xres_virtual);
-
-		if (destx > srcx) {
-			mod &= ~0x8000;
-			src += bpp * (width - 1);
-			dst += bpp * (width - 1);
-			pat += bpp * 2;
-		}
-		if (desty > srcy) {
-			mod &= ~0x4000;
-			src += bpp * (height - 1) * xres_virtual;
-			dst += bpp * (height - 1) * xres_virtual;
-			pat += bpp * 4;
-		}
-
-		*(acm + ACM_SOURCE/4) = cpu_to_le32(src);
-	}
-
-	*(acm + ACM_PATTERN/4) = cpu_to_le32(pat);
-
-	*(acm + ACM_DESTINATION/4) = cpu_to_le32(dst);
-
-	tmp = mod << 16;
-	*(acm + ACM_CONTROL/4) = tmp;
-
-	tmp  = width | (height << 16);
-	
-	*(acm + ACM_BITMAP_DIMENSION/4) = cpu_to_le32(tmp);
-
-	*(((volatile unsigned char *)acm) + ACM_START_STATUS) = 0x00;
-	*(((volatile unsigned char *)acm) + ACM_START_STATUS) = 0x01;
-
-	/*
-	 * No reason to wait for the blitter to finish, it is better
-	 * just to check if it has finished before we use it again.
-	 */
-#if 1
-#if 0
-	while ((*(((volatile unsigned char *)acm) +
-		  (ACM_START_STATUS + 2)) & 1) == 0);
-#else
-	do{
-		blt_status = *(((volatile unsigned char *)acm) +
-			       (ACM_START_STATUS + 2));
-	}
-	while ((blt_status & 1) == 0);
-#endif
-#endif
-}
+#include <linux/init.h>
 
-#if 0
-void retz3_fill (unsigned short x, unsigned short y, unsigned
-		 short width, unsigned short height,
-		 unsigned short mode, unsigned short color)
-{
-
-}
-#endif
-
-
-/**************************************************************
- * Move cursor to x, y
- */
-void retz3_MoveCursor (unsigned short x, unsigned short y)
-{
-	/* Guess we gotta deal with the cursor at some point */
-}
-
-
-/* -------------------- Interfaces to hardware functions -------------------- */
-
-
-static struct fb_hwswitch retz3_switch = {
-	retz3_init, retz3_encode_fix, retz3_decode_var, retz3_encode_var,
-	retz3_getcolreg, retz3_setcolreg, retz3_blank
-};
-
-
-/* -------------------- Generic routines ------------------------------------ */
-
-
-/*
- *    Fill the hardware's `par' structure.
- */
-
-static void retz3_fb_get_par(struct retz3_fb_par *par)
-{
-	if (current_par_valid)
-		*par = current_par;
-	else
-		fbhw->decode_var(&retz3_fb_predefined[z3fb_mode], par);
-}
-
-
-static void retz3_fb_set_par(struct retz3_fb_par *par)
-{
-	current_par = *par;
-	current_par_valid = 1;
-}
-
-
-static int do_fb_set_var(struct fb_var_screeninfo *var, int isactive)
-{
-	int err, activate;
-	struct retz3_fb_par par;
-
-	if ((err = fbhw->decode_var(var, &par)))
-		return err;
-	activate = var->activate;
-	if ((var->activate & FB_ACTIVATE_MASK) == FB_ACTIVATE_NOW && isactive)
-		retz3_fb_set_par(&par);
-	fbhw->encode_var(var, &par);
-	var->activate = activate;
-
-#if 1
-	retz3_set_video(var, &current_par);
-#endif
-	return 0;
-}
-
-
-/*
- *    Default Colormaps
- */
-
-static unsigned short red16[] =
-	{ 0x0000, 0x0000, 0x0000, 0x0000, 0xc000, 0xc000, 0xc000, 0xc000,
-	  0x8000, 0x0000, 0x0000, 0x0000, 0xffff, 0xffff, 0xffff, 0xffff };
-static unsigned short green16[] =
-	{ 0x0000, 0x0000, 0xc000, 0xc000, 0x0000, 0x0000, 0xc000, 0xc000,
-	  0x8000, 0x0000, 0xffff, 0xffff, 0x0000, 0x0000, 0xffff, 0xffff };
-static unsigned short blue16[] =
-	{ 0x0000, 0xc000, 0x0000, 0xc000, 0x0000, 0xc000, 0x0000, 0xc000,
-	  0x8000, 0xffff, 0x0000, 0xffff, 0x0000, 0xffff, 0x0000, 0xffff };
-
-
-static struct fb_cmap default_16_colors =
-   { 0, 16, red16, green16, blue16, NULL };
-
-
-static struct fb_cmap *get_default_colormap(int bpp)
-{
-	return &default_16_colors;
-}
-
-
-#define CNVT_TOHW(val,width)     ((((val)<<(width))+0x7fff-(val))>>16)
-#define CNVT_FROMHW(val,width)   (((width) ? ((((val)<<16)-(val)) / \
-                                              ((1<<(width))-1)) : 0))
-
-static int do_fb_get_cmap(struct fb_cmap *cmap, struct fb_var_screeninfo *var,
-                          int kspc)
-{
-	int i, start;
-	unsigned short *red, *green, *blue, *transp;
-	unsigned int hred, hgreen, hblue, htransp;
-
-	red = cmap->red;
-	green = cmap->green;
-	blue = cmap->blue;
-	transp = cmap->transp;
-	start = cmap->start;
-
-	if (start < 0)
-		return -EINVAL;
-	for (i = 0; i < cmap->len; i++) {
-		if (fbhw->getcolreg(start++, &hred, &hgreen, &hblue, &htransp))
-			return 0;
-		hred = CNVT_FROMHW(hred, var->red.length);
-		hgreen = CNVT_FROMHW(hgreen, var->green.length);
-		hblue = CNVT_FROMHW(hblue, var->blue.length);
-		htransp = CNVT_FROMHW(htransp, var->transp.length);
-		if (kspc) {
-			*red = hred;
-			*green = hgreen;
-			*blue = hblue;
-			if (transp)
-				*transp = htransp;
-		} else {
-			put_user(hred, red);
-			put_user(hgreen, green);
-			put_user(hblue, blue);
-			if (transp)
-				put_user(htransp, transp);
-		}
-		red++;
-		green++;
-		blue++;
-		if (transp)
-			transp++;
-	}
-	return 0;
-}
-
-
-static int do_fb_set_cmap(struct fb_cmap *cmap, struct fb_var_screeninfo *var,
-                          int kspc)
-{
-	int i, start;
-	unsigned short *red, *green, *blue, *transp;
-	unsigned int hred, hgreen, hblue, htransp;
-
-	red = cmap->red;
-	green = cmap->green;
-	blue = cmap->blue;
-	transp = cmap->transp;
-	start = cmap->start;
-
-	if (start < 0)
-		return -EINVAL;
-	for (i = 0; i < cmap->len; i++) {
-		if (kspc) {
-			hred = *red;
-			hgreen = *green;
-			hblue = *blue;
-			htransp = transp ? *transp : 0;
-		} else {
-			get_user(hred, red);
-			get_user(hgreen, green);
-			get_user(hblue, blue);
-			if (transp)
-				get_user(htransp, transp);
-			else
-				htransp = 0;
-		}
-		hred = CNVT_TOHW(hred, var->red.length);
-		hgreen = CNVT_TOHW(hgreen, var->green.length);
-		hblue = CNVT_TOHW(hblue, var->blue.length);
-		htransp = CNVT_TOHW(htransp, var->transp.length);
-		red++;
-		green++;
-		blue++;
-		if (transp)
-			transp++;
-		if (fbhw->setcolreg(start++, hred, hgreen, hblue, htransp))
-			return 0;
-	}
-	return 0;
-}
-
-
-static void do_install_cmap(int con)
-{
-	if (con != currcon)
-		return;
-	if (disp[con].cmap.len)
-		do_fb_set_cmap(&disp[con].cmap, &disp[con].var, 1);
-	else
-		do_fb_set_cmap(get_default_colormap(disp[con].var.bits_per_pixel),
-                                          &disp[con].var, 1);
-}
-
-
-static void memcpy_fs(int fsfromto, void *to, void *from, int len)
-{
-	switch (fsfromto) {
-	case 0:
-		memcpy(to, from, len);
-		return;
-	case 1:
-		copy_from_user(to, from, len);
-		return;
-	case 2:
-		copy_to_user(to, from, len);
-		return;
-	}
-}
-
-
-static void copy_cmap(struct fb_cmap *from, struct fb_cmap *to, int fsfromto)
-{
-	int size;
-	int tooff = 0, fromoff = 0;
-
-	if (to->start > from->start)
-		fromoff = to->start-from->start;
-	else
-		tooff = from->start-to->start;
-	size = to->len-tooff;
-	if (size > from->len-fromoff)
-		size = from->len-fromoff;
-	if (size < 0)
-		return;
-	size *= sizeof(unsigned short);
-	memcpy_fs(fsfromto, to->red+tooff, from->red+fromoff, size);
-	memcpy_fs(fsfromto, to->green+tooff, from->green+fromoff, size);
-	memcpy_fs(fsfromto, to->blue+tooff, from->blue+fromoff, size);
-	if (from->transp && to->transp)
-		memcpy_fs(fsfromto, to->transp+tooff,
-			  from->transp+fromoff, size);
-}
-
-
-static int alloc_cmap(struct fb_cmap *cmap, int len, int transp)
-{
-	int size = len*sizeof(unsigned short);
-
-	if (cmap->len != len) {
-		if (cmap->red)
-			kfree(cmap->red);
-		if (cmap->green)
-			kfree(cmap->green);
-		if (cmap->blue)
-			kfree(cmap->blue);
-		if (cmap->transp)
-			kfree(cmap->transp);
-		cmap->red = cmap->green = cmap->blue = cmap->transp = NULL;
-		cmap->len = 0;
-		if (!len)
-			return 0;
-		if (!(cmap->red = kmalloc(size, GFP_ATOMIC)))
-			return -1;
-		if (!(cmap->green = kmalloc(size, GFP_ATOMIC)))
-			return -1;
-		if (!(cmap->blue = kmalloc(size, GFP_ATOMIC)))
-			return -1;
-		if (transp) {
-			if (!(cmap->transp = kmalloc(size, GFP_ATOMIC)))
-				return -1;
-		} else
-			cmap->transp = NULL;
-	}
-	cmap->start = 0;
-	cmap->len = len;
-	copy_cmap(get_default_colormap(len), cmap, 0);
-	return 0;
-}
-
-
-/*
- *    Get the Fixed Part of the Display
- */
-
-static int retz3_fb_get_fix(struct fb_fix_screeninfo *fix, int con)
-{
-	struct retz3_fb_par par;
-	int error = 0;
-
-	if (con == -1)
-		retz3_fb_get_par(&par);
-	else
-		error = fbhw->decode_var(&disp[con].var, &par);
-	return(error ? error : fbhw->encode_fix(fix, &par));
-}
-
-
-/*
- *    Get the User Defined Part of the Display
- */
-
-static int retz3_fb_get_var(struct fb_var_screeninfo *var, int con)
-{
-	struct retz3_fb_par par;
-	int error = 0;
-
-	if (con == -1) {
-		retz3_fb_get_par(&par);
-		error = fbhw->encode_var(var, &par);
-	} else
-		*var = disp[con].var;
-	return error;
-}
-
-
-static void retz3_fb_set_disp(int con)
-{
-	struct fb_fix_screeninfo fix;
-
-	retz3_fb_get_fix(&fix, con);
-	if (con == -1)
-		con = 0;
-	disp[con].screen_base = (unsigned char *)fix.smem_start;
-	disp[con].visual = fix.visual;
-	disp[con].type = fix.type;
-	disp[con].type_aux = fix.type_aux;
-	disp[con].ypanstep = fix.ypanstep;
-	disp[con].ywrapstep = fix.ywrapstep;
-	disp[con].can_soft_blank = 1;
-	disp[con].inverse = z3fb_inverse;
-}
-
-
-/*
- *    Set the User Defined Part of the Display
- */
-
-static int retz3_fb_set_var(struct fb_var_screeninfo *var, int con)
-{
-	int err, oldxres, oldyres, oldvxres, oldvyres, oldbpp;
-
-	if ((err = do_fb_set_var(var, con == currcon)))
-		return err;
-	if ((var->activate & FB_ACTIVATE_MASK) == FB_ACTIVATE_NOW) {
-		oldxres = disp[con].var.xres;
-		oldyres = disp[con].var.yres;
-		oldvxres = disp[con].var.xres_virtual;
-		oldvyres = disp[con].var.yres_virtual;
-		oldbpp = disp[con].var.bits_per_pixel;
-		disp[con].var = *var;
-		if (oldxres != var->xres || oldyres != var->yres ||
-		    oldvxres != var->xres_virtual ||
-		    oldvyres != var->yres_virtual ||
-		    oldbpp != var->bits_per_pixel) {
-			retz3_fb_set_disp(con);
-			(*fb_info.changevar)(con);
-			alloc_cmap(&disp[con].cmap, 0, 0);
-			do_install_cmap(con);
-		}
-	}
-	var->activate = 0;
-	return 0;
-}
-
-
-/*
- *    Get the Colormap
- */
-
-static int retz3_fb_get_cmap(struct fb_cmap *cmap, int kspc, int con)
-{
-	if (con == currcon) /* current console? */
-		return(do_fb_get_cmap(cmap, &disp[con].var, kspc));
-	else if (disp[con].cmap.len) /* non default colormap? */
-		copy_cmap(&disp[con].cmap, cmap, kspc ? 0 : 2);
-	else
-		copy_cmap(get_default_colormap(disp[con].var.bits_per_pixel),
-			  cmap, kspc ? 0 : 2);
-	return 0;
-}
-
-
-/*
- *    Set the Colormap
- */
-
-static int retz3_fb_set_cmap(struct fb_cmap *cmap, int kspc, int con)
-{
-	int err;
-
-	if (!disp[con].cmap.len) {       /* no colormap allocated? */
-		if ((err = alloc_cmap(&disp[con].cmap,
-				      1<<disp[con].var.bits_per_pixel, 0)))
-			return err;
-	}
-	if (con == currcon)              /* current console? */
-		return(do_fb_set_cmap(cmap, &disp[con].var, kspc));
-	else
-		copy_cmap(cmap, &disp[con].cmap, kspc ? 0 : 1);
-	return 0;
-}
-
-
-/*
- *    Pan or Wrap the Display
- *
- *    This call looks only at xoffset, yoffset and the FB_VMODE_YWRAP flag
- */
-
-static int retz3_fb_pan_display(struct fb_var_screeninfo *var, int con)
-{
-	return -EINVAL;
-}
-
-
-/*
- *    RetinaZ3 Frame Buffer Specific ioctls
- */
-
-static int retz3_fb_ioctl(struct inode *inode, struct file *file,
-                          unsigned int cmd, unsigned long arg, int con)
-{
-	return -EINVAL;
-}
-
-
-static struct fb_ops retz3_fb_ops = {
-   retz3_fb_get_fix, retz3_fb_get_var, retz3_fb_set_var, retz3_fb_get_cmap,
-   retz3_fb_set_cmap, retz3_fb_pan_display, retz3_fb_ioctl
-};
-
-
-int retz3_probe(void)
-{
-	z3_key = zorro_find(MANUF_MACROSYSTEMS2, PROD_RETINA_Z3, 0, 0);
-	return(z3_key);
-}
-
-
-void retz3_video_setup(char *options, int *ints)
-{
-	char *this_opt;
-	int i;
-
-	fb_info.fontname[0] = '\0';
-
-	if (!options || !*options)
-		return;
-
-	for (this_opt = strtok(options, ","); this_opt; 
-	     this_opt = strtok(NULL, ",")){
-		if (!strcmp(this_opt, "inverse")) {
-			z3fb_inverse = 1;
-			for (i = 0; i < 16; i++) {
-				red16[i] = ~red16[i];
-				green16[i] = ~green16[i];
-				blue16[i] = ~blue16[i];
-			}
-		} else if (!strncmp(this_opt, "font:", 5))
-			strcpy(fb_info.fontname, this_opt+5);
-		else
-			z3fb_mode = get_video_mode(this_opt);
-	}
-}
-
-
-/*
- *    Initialization
- */
-
-struct fb_info *retz3_fb_init(long *mem_start)
-{
-	int err;
-	struct retz3_fb_par par;
-
-	memstart = mem_start;
-
-	fbhw = &retz3_switch;
-
-	err = register_framebuffer(retz3_fb_name, &node, &retz3_fb_ops,
-				   NUM_TOTAL_MODES, retz3_fb_predefined);
-	if (err < 0)
-		panic("Cannot register frame buffer\n");
-
-	fbhw->init();
-
-	if (z3fb_mode == -1)
-		z3fb_mode = 1;
-
-	fbhw->decode_var(&retz3_fb_predefined[z3fb_mode], &par);
-	fbhw->encode_var(&retz3_fb_predefined[0], &par);
-
-	strcpy(fb_info.modename, retz3_fb_name);
-	fb_info.disp = disp;
-	fb_info.switch_con = &z3fb_switch;
-	fb_info.updatevar = &z3fb_updatevar;
-	fb_info.blank = &z3fb_blank;
-	fb_info.setcmap = &z3fb_setcmap;
-
-	do_fb_set_var(&retz3_fb_predefined[0], 0);
-	retz3_fb_get_var(&disp[0].var, -1);
-	retz3_fb_set_disp(-1);
-	do_install_cmap(0);
-
-	return &fb_info;
-}
-
-
-static int z3fb_switch(int con)
-{
-	/* Do we have to save the colormap? */
-	if (disp[currcon].cmap.len)
-		do_fb_get_cmap(&disp[currcon].cmap, &disp[currcon].var, 1);
-
-	do_fb_set_var(&disp[con].var, 1);
-	currcon = con;
-	/* Install new colormap */
-	do_install_cmap(con);
-	return 0;
-}
-
-
-/*
- *    Update the `var' structure (called by fbcon.c)
- *
- *    This call looks only at yoffset and the FB_VMODE_YWRAP flag in `var'.
- *    Since it's called by a kernel driver, no range checking is done.
- */
-
-static int z3fb_updatevar(int con)
-{
-	return 0;
-}
-
-
-/*
- *    Blank the display.
- */
-
-static void z3fb_blank(int blank)
-{
-	fbhw->blank(blank);
-}
-
-
-/*
- *    Set the colormap
- */
-
-static int z3fb_setcmap(struct fb_cmap *cmap, int con)
-{
-	return(retz3_fb_set_cmap(cmap, 1, con));
-}
-
-
-/*
- *    Get a Video Mode
- */
-
-static int get_video_mode(const char *name)
-{
-	int i;
-
-	for (i = 1; i <= NUM_PREDEF_MODES; i++)
-		if (!strcmp(name, retz3_fb_modenames[i])){
-			retz3_fb_predefined[0] = retz3_fb_predefined[i];
-			return i;
-		}
-	return -1;
-}
-/*
- * Linux/arch/m68k/amiga/retz3fb.c -- Low level implementation of the
- *                                    RetinaZ3 frame buffer device
- *
- *    Copyright (C) 1997 Jes Sorensen
- *
- * This file is based on the CyberVision64 frame buffer device and
- * the generic Cirrus Logic driver.
- *
- * cyberfb.c: Copyright (C) 1996 Martin Apel,
- *                               Geert Uytterhoeven
- * clgen.c:   Copyright (C) 1996 Frank Neumann
- *
- * History:
- *   - 22 Jan 97: Initial work
- *   - 14 Feb 97: Screen initialization works somewhat, still only
- *                8-bit packed pixel is supported.
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file COPYING in the main directory of this archive
- * for more details.
- */
-
-
-#include <linux/kernel.h>
-#include <linux/errno.h>
-#include <linux/string.h>
-#include <linux/mm.h>
-#include <linux/tty.h>
-#include <linux/malloc.h>
-#include <linux/delay.h>
-#include <linux/fb.h>
 #include <asm/uaccess.h>
 #include <asm/system.h>
 #include <asm/irq.h>
-#include <linux/zorro.h>
 #include <asm/pgtable.h>
 
 #include "retz3fb.h"
@@ -3402,7 +1652,7 @@ void retz3_video_setup(char *options, int *ints)
  *    Initialization
  */
 
-struct fb_info *retz3_fb_init(long *mem_start)
+__initfunc(struct fb_info *retz3_fb_init(long *mem_start))
 {
 	int err;
 	struct retz3_fb_par par;
diff --git a/arch/m68k/amiga/retz3fb.h b/arch/m68k/amiga/retz3fb.h
index 9ff209060..0cb03c324 100644
--- a/arch/m68k/amiga/retz3fb.h
+++ b/arch/m68k/amiga/retz3fb.h
@@ -284,289 +284,3 @@
 #define Z3BLTorInverted		0xb0	/* NOT src OR dst */
 #define Z3BLTnand		0x70	/* NOT src OR NOT dst */
 #define Z3BLTset		0xf0	/* 1 */
-/*
- * Linux/arch/m68k/amiga/retz3fb.h -- Defines and macros for the
- *                                    RetinaZ3 frame buffer device
- *
- *    Copyright (C) 1997 Jes Sorensen
- *
- * History:
- *   - 22 Jan 97: Initial work
- *
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file COPYING in the main directory of this archive
- * for more details.
- */
-
-/*
- * Macros to read and write to registers.
- */
-#define reg_w(reg,dat) (*(z3_regs + reg) = dat)
-#define reg_r(reg) (*(z3_regs + reg))
-
-/*
- * Macro to access the sequencer.
- */
-#define seq_w(sreg,sdat) \
-	do{ reg_w(SEQ_IDX, sreg); reg_w(SEQ_DATA, sdat); } while(0)
-
-/*
- * Macro to access the CRT controller.
- */
-#define crt_w(creg,cdat) \
-	do{ reg_w(CRT_IDX, creg); reg_w(CRT_DATA, cdat); } while(0)
-
-/*
- * Macro to access the graphics controller.
- */
-#define gfx_w(greg,gdat) \
-	do{ reg_w(GFX_IDX, greg); reg_w(GFX_DATA, gdat); } while(0)
-
-/*
- * Macro to access the attribute controller.
- */
-#define attr_w(areg,adat) \
-	do{ reg_w(ACT_IDX, areg); reg_w(ACT_DATA, adat); } while(0)
-
-/*
- * Macro to access the pll.
- */
-#define pll_w(preg,pdat) \
-	do{ reg_w(PLL_IDX, preg); \
-	    reg_w(PLL_DATA, (pdat & 0xff)); \
-	    reg_w(PLL_DATA, (pdat >> 8));\
-	} while(0)
-
-/*
- * Offsets
- */
-#define VIDEO_MEM_OFFSET	0x00c00000
-#define ACM_OFFSET		0x00b00000
-
-/*
- * Accelerator Control Menu
- */
-#define ACM_PRIMARY_OFFSET	0x00
-#define ACM_SECONDARY_OFFSET	0x04
-#define ACM_MODE_CONTROL	0x08
-#define ACM_CURSOR_POSITION	0x0c
-#define ACM_START_STATUS	0x30
-#define ACM_CONTROL		0x34
-#define ACM_RASTEROP_ROTATION	0x38
-#define ACM_BITMAP_DIMENSION	0x3c
-#define ACM_DESTINATION		0x40
-#define ACM_SOURCE		0x44
-#define ACM_PATTERN		0x48
-#define ACM_FOREGROUND		0x4c
-#define ACM_BACKGROUND		0x50
-
-/*
- * Video DAC addresses
- */
-#define VDAC_ADDRESS		0x03c8
-#define VDAC_ADDRESS_W		0x03c8
-#define VDAC_ADDRESS_R		0x03c7
-#define VDAC_STATE		0x03c7
-#define VDAC_DATA		0x03c9
-#define VDAC_MASK		0x03c6
-
-/*
- * Sequencer
- */
-#define SEQ_IDX			0x03c4	/* Sequencer Index */
-#define SEQ_DATA		0x03c5
-#define SEQ_RESET		0x00
-#define SEQ_CLOCKING_MODE	0x01
-#define SEQ_MAP_MASK		0x02
-#define SEQ_CHAR_MAP_SELECT	0x03
-#define SEQ_MEMORY_MODE		0x04
-#define SEQ_EXTENDED_ENABLE	0x05	/* NCR extensions */
-#define SEQ_UNKNOWN1         	0x06
-#define SEQ_UNKNOWN2         	0x07
-#define SEQ_CHIP_ID		0x08
-#define SEQ_UNKNOWN3         	0x09
-#define SEQ_CURSOR_COLOR1	0x0a
-#define SEQ_CURSOR_COLOR0	0x0b
-#define SEQ_CURSOR_CONTROL	0x0c
-#define SEQ_CURSOR_X_LOC_HI	0x0d
-#define SEQ_CURSOR_X_LOC_LO	0x0e
-#define SEQ_CURSOR_Y_LOC_HI	0x0f
-#define SEQ_CURSOR_Y_LOC_LO	0x10
-#define SEQ_CURSOR_X_INDEX	0x11
-#define SEQ_CURSOR_Y_INDEX	0x12
-#define SEQ_CURSOR_STORE_HI	0x13
-#define SEQ_CURSOR_STORE_LO	0x14
-#define SEQ_CURSOR_ST_OFF_HI	0x15
-#define SEQ_CURSOR_ST_OFF_LO	0x16
-#define SEQ_CURSOR_PIXELMASK	0x17
-#define SEQ_PRIM_HOST_OFF_HI	0x18
-#define SEQ_PRIM_HOST_OFF_LO	0x19
-#define SEQ_LINEAR_0		0x1a
-#define SEQ_LINEAR_1		0x1b
-#define SEQ_SEC_HOST_OFF_HI	0x1c
-#define SEQ_SEC_HOST_OFF_LO	0x1d
-#define SEQ_EXTENDED_MEM_ENA	0x1e
-#define SEQ_EXT_CLOCK_MODE	0x1f
-#define SEQ_EXT_VIDEO_ADDR	0x20
-#define SEQ_EXT_PIXEL_CNTL	0x21
-#define SEQ_BUS_WIDTH_FEEDB	0x22
-#define SEQ_PERF_SELECT		0x23
-#define SEQ_COLOR_EXP_WFG	0x24
-#define SEQ_COLOR_EXP_WBG	0x25
-#define SEQ_EXT_RW_CONTROL	0x26
-#define SEQ_MISC_FEATURE_SEL	0x27
-#define SEQ_COLOR_KEY_CNTL	0x28
-#define SEQ_COLOR_KEY_MATCH0	0x29
-#define SEQ_COLOR_KEY_MATCH1 	0x2a
-#define SEQ_COLOR_KEY_MATCH2 	0x2b
-#define SEQ_UNKNOWN6         	0x2c
-#define SEQ_CRC_CONTROL		0x2d
-#define SEQ_CRC_DATA_LOW	0x2e
-#define SEQ_CRC_DATA_HIGH	0x2f
-#define SEQ_MEMORY_MAP_CNTL	0x30
-#define SEQ_ACM_APERTURE_1	0x31
-#define SEQ_ACM_APERTURE_2	0x32
-#define SEQ_ACM_APERTURE_3	0x33
-#define SEQ_BIOS_UTILITY_0	0x3e
-#define SEQ_BIOS_UTILITY_1	0x3f
-
-/*
- * Graphics Controller
- */
-#define GFX_IDX			0x03ce
-#define GFX_DATA		0x03cf
-#define GFX_SET_RESET		0x00
-#define GFX_ENABLE_SET_RESET	0x01
-#define GFX_COLOR_COMPARE	0x02
-#define GFX_DATA_ROTATE		0x03
-#define GFX_READ_MAP_SELECT	0x04
-#define GFX_GRAPHICS_MODE	0x05
-#define GFX_MISC		0x06
-#define GFX_COLOR_XCARE		0x07
-#define GFX_BITMASK		0x08
-
-/*
- * CRT Controller
- */
-#define CRT_IDX			0x03d4
-#define CRT_DATA		0x03d5
-#define CRT_HOR_TOTAL		0x00
-#define CRT_HOR_DISP_ENA_END	0x01
-#define CRT_START_HOR_BLANK	0x02
-#define CRT_END_HOR_BLANK	0x03
-#define CRT_START_HOR_RETR	0x04
-#define CRT_END_HOR_RETR	0x05
-#define CRT_VER_TOTAL		0x06
-#define CRT_OVERFLOW		0x07
-#define CRT_PRESET_ROW_SCAN	0x08
-#define CRT_MAX_SCAN_LINE	0x09
-#define CRT_CURSOR_START	0x0a
-#define CRT_CURSOR_END		0x0b
-#define CRT_START_ADDR_HIGH	0x0c
-#define CRT_START_ADDR_LOW	0x0d
-#define CRT_CURSOR_LOC_HIGH	0x0e
-#define CRT_CURSOR_LOC_LOW	0x0f
-#define CRT_START_VER_RETR	0x10
-#define CRT_END_VER_RETR	0x11
-#define CRT_VER_DISP_ENA_END	0x12
-#define CRT_OFFSET		0x13
-#define CRT_UNDERLINE_LOC	0x14
-#define CRT_START_VER_BLANK	0x15
-#define CRT_END_VER_BLANK	0x16
-#define CRT_MODE_CONTROL	0x17
-#define CRT_LINE_COMPARE	0x18
-#define CRT_UNKNOWN1         	0x19
-#define CRT_UNKNOWN2         	0x1a
-#define CRT_UNKNOWN3         	0x1b
-#define CRT_UNKNOWN4         	0x1c
-#define CRT_UNKNOWN5         	0x1d
-#define CRT_UNKNOWN6         	0x1e
-#define CRT_UNKNOWN7         	0x1f
-#define CRT_UNKNOWN8         	0x20
-#define CRT_UNKNOWN9		0x21
-#define CRT_UNKNOWN10		0x22
-#define CRT_UNKNOWN11      	0x23
-#define CRT_UNKNOWN12      	0x24
-#define CRT_UNKNOWN13      	0x25
-#define CRT_UNKNOWN14      	0x26
-#define CRT_UNKNOWN15      	0x27
-#define CRT_UNKNOWN16      	0x28
-#define CRT_UNKNOWN17      	0x29
-#define CRT_UNKNOWN18      	0x2a
-#define CRT_UNKNOWN19      	0x2b
-#define CRT_UNKNOWN20      	0x2c
-#define CRT_UNKNOWN21      	0x2d
-#define CRT_UNKNOWN22      	0x2e
-#define CRT_UNKNOWN23      	0x2f
-#define CRT_EXT_HOR_TIMING1	0x30	/* NCR crt extensions */
-#define CRT_EXT_START_ADDR	0x31
-#define CRT_EXT_HOR_TIMING2	0x32
-#define CRT_EXT_VER_TIMING	0x33
-#define CRT_MONITOR_POWER	0x34
-
-/*
- * General Registers
- */
-#define GREG_STATUS0_R		0x03c2
-#define GREG_STATUS1_R		0x03da
-#define GREG_MISC_OUTPUT_R	0x03cc
-#define GREG_MISC_OUTPUT_W	0x03c2	
-#define GREG_FEATURE_CONTROL_R	0x03ca
-#define GREG_FEATURE_CONTROL_W	0x03da
-#define GREG_POS		0x0102
-
-/*
- * Attribute Controller
- */
-#define ACT_IDX			0x03C0
-#define ACT_ADDRESS_R		0x03C0
-#define ACT_DATA		0x03C0
-#define ACT_ADDRESS_RESET	0x03DA
-#define ACT_PALETTE0		0x00
-#define ACT_PALETTE1		0x01
-#define ACT_PALETTE2		0x02
-#define ACT_PALETTE3		0x03
-#define ACT_PALETTE4		0x04
-#define ACT_PALETTE5		0x05
-#define ACT_PALETTE6		0x06
-#define ACT_PALETTE7		0x07
-#define ACT_PALETTE8		0x08
-#define ACT_PALETTE9		0x09
-#define ACT_PALETTE10		0x0A
-#define ACT_PALETTE11		0x0B
-#define ACT_PALETTE12		0x0C
-#define ACT_PALETTE13		0x0D
-#define ACT_PALETTE14		0x0E
-#define ACT_PALETTE15		0x0F
-#define ACT_ATTR_MODE_CNTL	0x10
-#define ACT_OVERSCAN_COLOR	0x11
-#define ACT_COLOR_PLANE_ENA	0x12
-#define ACT_HOR_PEL_PANNING	0x13
-#define ACT_COLOR_SELECT	0x14
-
-/*
- * PLL
- */
-#define PLL_IDX			0x83c8
-#define PLL_DATA		0x83c9
-
-/*
- * Blitter operations
- */
-#define	Z3BLTclear		0x00	/* 0 */
-#define Z3BLTand		0x80	/* src AND dst */
-#define Z3BLTandReverse		0x40	/* src AND NOT dst */
-#define Z3BLTcopy		0xc0	/* src */
-#define Z3BLTandInverted	0x20	/* NOT src AND dst */
-#define	Z3BLTnoop		0xa0	/* dst */
-#define Z3BLTxor		0x60	/* src XOR dst */
-#define Z3BLTor			0xe0	/* src OR dst */
-#define Z3BLTnor		0x10	/* NOT src AND NOT dst */
-#define Z3BLTequiv		0x90	/* NOT src XOR dst */
-#define Z3BLTinvert		0x50	/* NOT dst */
-#define Z3BLTorReverse		0xd0	/* src OR NOT dst */
-#define Z3BLTcopyInverted	0x30	/* NOT src */
-#define Z3BLTorInverted		0xb0	/* NOT src OR dst */
-#define Z3BLTnand		0x70	/* NOT src OR NOT dst */
-#define Z3BLTset		0xf0	/* 1 */
diff --git a/arch/m68k/amiga/zorro.c b/arch/m68k/amiga/zorro.c
index acca3e6bc..d64ad8ce3 100644
--- a/arch/m68k/amiga/zorro.c
+++ b/arch/m68k/amiga/zorro.c
@@ -13,10 +13,11 @@
 #include <linux/types.h>
 #include <linux/kernel.h>
 #include <linux/string.h>
+#include <linux/init.h>
+#include <linux/zorro.h>
 #include <asm/setup.h>
 #include <asm/bitops.h>
 #include <asm/amigahw.h>
-#include <linux/zorro.h>
 
 
 #ifdef CONFIG_ZORRO
@@ -415,10 +416,6 @@ BEGIN_PROD(MASOBOSHI)
    PROD("MVD 819", MVD_819)
 END
 
-BEGIN_PROD(DELACOMP)
-   PROD("RAM Expansion 2000", DELACOMP_RAM_2000)
-END
-
 BEGIN_PROD(VILLAGE_TRONIC)
    PROD("Domino Graphics Board (RAM)", DOMINO_RAM)
    PROD("Domino Graphics Board (REG)", DOMINO_REG)
@@ -656,7 +653,6 @@ BEGIN_MANUF
    MANUF("Helfrich", HELFRICH1)
    MANUF("Software Result Enterprises", SW_RESULT_ENTS)
    MANUF("Masoboshi", MASOBOSHI)
-   MANUF("DelaComp", DELACOMP)
    MANUF("Village Tronic", VILLAGE_TRONIC)
    MANUF("Utilities Unlimited", UTILITIES_ULTD)
    MANUF("Amitrix", AMITRIX)
@@ -992,7 +988,7 @@ static void mark_region(u_long addr, u_long size, int flag)
     *    Initialization
     */
 
-void zorro_init(void)
+__initfunc(void zorro_init(void))
 {
    int i;
    struct ConfigDev *cd;
diff --git a/arch/m68k/atari/atafb.c b/arch/m68k/atari/atafb.c
index 64c9a62d3..efa6fc8b8 100644
--- a/arch/m68k/atari/atafb.c
+++ b/arch/m68k/atari/atafb.c
@@ -46,6 +46,7 @@
 #include <linux/tty.h>
 #include <linux/malloc.h>
 #include <linux/delay.h>
+#include <linux/init.h>
 
 #include <asm/setup.h>
 #include <asm/uaccess.h>
@@ -2911,8 +2912,8 @@ atafb_setcmap(struct fb_cmap *cmap, int con)
 	return(atari_fb_set_cmap(cmap, 1, con));
 }
 
-struct fb_info *
-atari_fb_init(long *mem_start)
+__initfunc(struct fb_info *
+atari_fb_init(long *mem_start))
 {
 	int err;
 	int pad;
diff --git a/arch/m68k/atari/ataints.c b/arch/m68k/atari/ataints.c
index b65beefb1..b2887f7da 100644
--- a/arch/m68k/atari/ataints.c
+++ b/arch/m68k/atari/ataints.c
@@ -39,6 +39,7 @@
 #include <linux/kernel.h>
 #include <linux/ptrace.h>
 #include <linux/kernel_stat.h>
+#include <linux/init.h>
 
 #include <asm/system.h>
 #include <asm/traps.h>
@@ -162,15 +163,19 @@ static int free_vme_vec_bitmap = 0;
 
 #define	MFP_MK_BASE	"0xfa13"
 
-/* This must agree with head.S.  */
-#define ORIG_DO "0x20"
-#define FORMATVEC "0x2E"
-#define SR "0x28"
+/* This must agree with entry.S.  */
+#define ORIG_DO "0x24"
+#define FORMATVEC "0x32"
+#define SR "0x2C"
 #define SAVE_ALL				\
 	"clrl	%%sp@-;"    /* stk_adj */	\
 	"pea	-1:w;"	    /* orig d0 = -1 */	\
 	"movel	%%d0,%%sp@-;" /* d0 */		\
-	"moveml	%%d1-%%d5/%%a0-%%a1,%%sp@-"
+	"moveml	%%d1-%%d5/%%a0-%%a2,%%sp@-"
+#define GET_CURRENT(tmp) \
+	"movel	%%sp,"#tmp";" \
+	"andw	#-8192,"#tmp";" \
+	"movel	"#tmp",%%a2"
 
 #define	BUILD_SLOW_IRQ(n)						   \
 asmlinkage void IRQ_NAME(n);						   \
@@ -180,6 +185,7 @@ __asm__ (ALIGN_STR "\n"							   \
 SYMBOL_NAME_STR(atari_slow_irq_) #n "_handler:\t"			   \
 "	addql	#1,"SYMBOL_NAME_STR(local_irq_count)"\n"		   \
 	SAVE_ALL "\n"							   \
+	GET_CURRENT(%%d0) "\n"						   \
 "	andb	#~(1<<(" #n "&7)),"	/* mask this interrupt */	   \
 	"("MFP_MK_BASE"+(((" #n "&8)^8)>>2)+((" #n "&16)<<3)):w\n"	   \
 "	bfextu	%%sp@("SR"){#5,#3},%%d0\n" /* get old IPL from stack frame */ \
@@ -282,7 +288,8 @@ SYMBOL_NAME_STR(atari_fast_irq_handler) ":
 	orw 	#0x700,%%sr		/* disable all interrupts */
 "SYMBOL_NAME_STR(atari_prio_irq_handler) ":\t
 	addql	#1,"SYMBOL_NAME_STR(local_irq_count)"\n"
-	SAVE_ALL "
+	SAVE_ALL "\n"
+	GET_CURRENT(%%d0) "
 	/* get vector number from stack frame and convert to source */
 	bfextu	%%sp@(" FORMATVEC "){#4,#10},%%d0
 	subw	#(0x40-8),%%d0
@@ -331,7 +338,7 @@ extern void atari_microwire_cmd( int cmd );
  * the atari IRQ handling routines.
  */
 
-void atari_init_IRQ(void)
+__initfunc(void atari_init_IRQ(void))
 {
 	int i;
 
diff --git a/arch/m68k/atari/atakeyb.c b/arch/m68k/atari/atakeyb.c
index 4f57cb928..beccf9a84 100644
--- a/arch/m68k/atari/atakeyb.c
+++ b/arch/m68k/atari/atakeyb.c
@@ -22,6 +22,7 @@
 #include <linux/timer.h>
 #include <linux/kd.h>
 #include <linux/random.h>
+#include <linux/init.h>
 
 #include <asm/atariints.h>
 #include <asm/atarihw.h>
@@ -804,7 +805,7 @@ void atari_kbd_leds (unsigned int leds)
  * Martin Rogge, 20 Aug 1995
  */
  
-int atari_keyb_init(void)
+__initfunc(int atari_keyb_init(void))
 {
     /* setup key map */
     key_maps[0]  = ataplain_map;
diff --git a/arch/m68k/atari/config.c b/arch/m68k/atari/config.c
index 0d0607da0..3cc2a840a 100644
--- a/arch/m68k/atari/config.c
+++ b/arch/m68k/atari/config.c
@@ -32,6 +32,7 @@
 #include <linux/tty.h>
 #include <linux/console.h>
 #include <linux/interrupt.h>
+#include <linux/init.h>
 
 #include <asm/bootinfo.h>
 #include <asm/setup.h>
@@ -106,7 +107,7 @@ extern void (*kd_mksound)(unsigned int, unsigned int);
  * a temporary VBR and a vector table for the duration of the test.
  */
 
-static int hwreg_present( volatile void *regp )
+__initfunc(static int hwreg_present( volatile void *regp ))
 {
     int	ret = 0;
     long	save_sp, save_vbr;
@@ -132,9 +133,8 @@ static int hwreg_present( volatile void *regp )
 }
   
 #if 0
-static int hwreg_present_bywrite( volatile void *regp,
-				 unsigned char val )
-
+__initfunc(static int
+hwreg_present_bywrite(volatile void *regp, unsigned char val))
 {
     int		ret;
     long	save_sp, save_vbr;
@@ -166,7 +166,7 @@ static int hwreg_present_bywrite( volatile void *regp,
 /* Basically the same, but writes a value into a word register, protected
  * by a bus error handler */
 
-static int hwreg_write( volatile void *regp, unsigned short val )
+__initfunc(static int hwreg_write( volatile void *regp, unsigned short val ))
 {
 	int		ret;
 	long	save_sp, save_vbr;
@@ -201,7 +201,7 @@ static int hwreg_write( volatile void *regp, unsigned short val )
  * should be readable without trouble (from channel A!).
  */
 
-static int scc_test( volatile char *ctla )
+__initfunc(static int scc_test( volatile char *ctla ))
 {
 	if (!hwreg_present( ctla ))
 		return( 0 );
@@ -228,7 +228,7 @@ static int scc_test( volatile char *ctla )
      *  Parse an Atari-specific record in the bootinfo
      */
 
-int atari_parse_bootinfo(const struct bi_record *record)
+__initfunc(int atari_parse_bootinfo(const struct bi_record *record))
 {
     int unknown = 0;
     const u_long *data = record->data;
@@ -247,7 +247,7 @@ int atari_parse_bootinfo(const struct bi_record *record)
      *  Setup the Atari configuration info
      */
 
-void config_atari(void)
+__initfunc(void config_atari(void))
 {
     memset(&atari_hw_present, 0, sizeof(atari_hw_present));
 
@@ -493,7 +493,8 @@ void config_atari(void)
     }
 }
 
-static void atari_sched_init(void (*timer_routine)(int, void *, struct pt_regs *))
+__initfunc(static void
+atari_sched_init(void (*timer_routine)(int, void *, struct pt_regs *)))
 {
     /* set Timer C data Register */
     mfp.tim_dt_c = INT_TICKS;
@@ -976,7 +977,7 @@ static void atari_par_console_write (const char *str, unsigned int count)
 }
 
 
-static void atari_debug_init(void)
+__initfunc(static void atari_debug_init(void))
 {
 #ifdef CONFIG_KGDB
 	/* if the m68k_debug_device is used by the GDB stub, do nothing here */
diff --git a/arch/m68k/atari/joystick.c b/arch/m68k/atari/joystick.c
index de6014171..2e33b151f 100644
--- a/arch/m68k/atari/joystick.c
+++ b/arch/m68k/atari/joystick.c
@@ -11,6 +11,7 @@
 #include <linux/errno.h>
 #include <linux/major.h>
 #include <linux/poll.h>
+#include <linux/init.h>
 
 #include <asm/atarikb.h>
 #include <asm/atari_joystick.h>
@@ -128,7 +129,7 @@ struct file_operations atari_joystick_fops = {
 	release_joystick
 };
 
-int atari_joystick_init(void)
+__initfunc(int atari_joystick_init(void))
 {
     joystick[0].active = joystick[1].active = 0;
     joystick[0].ready = joystick[1].ready = 0;
diff --git a/arch/m68k/atari/stdma.c b/arch/m68k/atari/stdma.c
index 247b9f684..9dc82de18 100644
--- a/arch/m68k/atari/stdma.c
+++ b/arch/m68k/atari/stdma.c
@@ -32,6 +32,8 @@
 #include <linux/types.h>
 #include <linux/genhd.h>
 #include <linux/sched.h>
+#include <linux/init.h>
+
 #include <asm/atari_stdma.h>
 #include <asm/atariints.h>
 #include <asm/atarihw.h>
@@ -171,7 +173,7 @@ int stdma_islocked(void)
  *
  */
 
-void stdma_init(void)
+__initfunc(void stdma_init(void))
 {
 	stdma_isr = NULL;
 	request_irq(IRQ_MFP_FDC, stdma_int, IRQ_TYPE_SLOW,
diff --git a/arch/m68k/atari/stram.c b/arch/m68k/atari/stram.c
index 565b20535..4e5f57bc9 100644
--- a/arch/m68k/atari/stram.c
+++ b/arch/m68k/atari/stram.c
@@ -149,6 +149,7 @@ atari_stram_free (void *ptr)
 #else
 
 #include <linux/mm.h>
+#include <linux/init.h>
 
 /* ++roman:
  * 
@@ -190,7 +191,7 @@ static unsigned long stram_end;
     /* Overall end of ST-Ram */
 
 
-void atari_stram_init( void )
+__initfunc(void atari_stram_init( void ))
 
 {	int		i;
 
diff --git a/arch/m68k/boot/amiga/linuxboot.c b/arch/m68k/boot/amiga/linuxboot.c
index d35966686..23b7fa9d0 100644
--- a/arch/m68k/boot/amiga/linuxboot.c
+++ b/arch/m68k/boot/amiga/linuxboot.c
@@ -22,6 +22,8 @@
  *  for more details.
  *
  *  History:
+ *	27 Mar 1997 FPU-less machines couldn't boot kernels that use bootinfo
+ *		    interface version 1.0 (Geert)
  *	03 Feb 1997 Implemented kernel decompression (Geert, based on Roman's
  *		    code for ataboot)
  *	30 Dec 1996 Reverted the CPU detection to the old scheme
@@ -791,6 +793,7 @@ static u_long get_chipset(void)
 static void get_processor(u_long *cpu, u_long *fpu, u_long *mmu)
 {
     *cpu = *fpu = 0;
+
     if (SysBase->AttnFlags & AFF_68060)
 	*cpu = CPU_68060;
     else if (SysBase->AttnFlags & AFF_68040)
@@ -799,15 +802,15 @@ static void get_processor(u_long *cpu, u_long *fpu, u_long *mmu)
 	*cpu = CPU_68030;
     else if (SysBase->AttnFlags & AFF_68020)
 	*cpu = CPU_68020;
+
     if (*cpu == CPU_68040 || *cpu == CPU_68060) {
 	if (SysBase->AttnFlags & AFF_FPU40)
 	    *fpu = *cpu;
-    } else {
-	if (SysBase->AttnFlags & AFF_68882)
-	    *fpu = FPU_68882;
-	else if (SysBase->AttnFlags & AFF_68881)
-	    *fpu = FPU_68881;
-    }
+    } else if (SysBase->AttnFlags & AFF_68882)
+	*fpu = FPU_68882;
+    else if (SysBase->AttnFlags & AFF_68881)
+	*fpu = FPU_68881;
+
     *mmu = *cpu;
 }
 
@@ -1047,7 +1050,7 @@ static int create_compat_bootinfo(void)
 	compat_bootinfo.cputype |= COMPAT_FPU_68040;
     else if (bi.fputype & FPU_68060)
 	compat_bootinfo.cputype |= COMPAT_FPU_68060;
-    else {
+    else if (bi.fputype) {
 	Printf("FPU type 0x%08lx not supported by kernel\n", bi.fputype);
 	return(0);
     }
diff --git a/arch/m68k/boot/amiga/linuxboot.h b/arch/m68k/boot/amiga/linuxboot.h
index 8ebfe63b4..e04425a3a 100644
--- a/arch/m68k/boot/amiga/linuxboot.h
+++ b/arch/m68k/boot/amiga/linuxboot.h
@@ -31,7 +31,7 @@
      *  Amiboot Version
      */
 
-#define AMIBOOT_VERSION		"5.4"
+#define AMIBOOT_VERSION		"5.5"
 
 
     /*
diff --git a/arch/m68k/fpsp040/skeleton.S b/arch/m68k/fpsp040/skeleton.S
index 5768516bd..466cf3fbc 100644
--- a/arch/m68k/fpsp040/skeleton.S
+++ b/arch/m68k/fpsp040/skeleton.S
@@ -51,13 +51,23 @@
  
 	.include "fpsp.h"
 
-LOFF_ORIG_D0	= 0x20
+/*
+ * This has to match entry.S
+ */
+LOFF_ORIG_D0	= 0x24
+
+#define curptr a2
 
 #define SAVE_ALL				\
 	clrl	%sp@-;     /* stk_adj */	\
 	movel	%d0,%sp@-; /* orig d0 */	\
 	movel	%d0,%sp@-; /* d0 */		\
-	moveml	%d1-%d5/%a0-%a1,%sp@-
+	moveml	%d1-%d5/%a0-%a1/%curptr,%sp@-;
+
+#define GET_CURRENT(tmp) \
+	movel	%sp,tmp; \
+	andw	&-8192,tmp; \
+	movel	tmp,%curptr;
 
 	|xref	b1238_fix
 
@@ -81,6 +91,7 @@ real_dz:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -175,6 +186,7 @@ inex_done:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -204,6 +216,7 @@ ovfl_done:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -233,6 +246,7 @@ unfl_done:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -258,6 +272,7 @@ real_snan:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -283,6 +298,7 @@ real_operr:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -314,6 +330,7 @@ real_bsun:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -338,6 +355,7 @@ real_fline:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -363,6 +381,7 @@ real_unsupp:
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -419,6 +438,7 @@ Lmustsched:
 	SAVE_ALL
 	moveq	#-1,%d0
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | indicate stack frame not for syscall
+	GET_CURRENT(%d0)
 	bral	SYMBOL_NAME(ret_from_exception)	| deliver signals, reschedule etc..
 
 
@@ -497,7 +517,7 @@ user_read:
 	movel	%d0,-(%sp)
 	movel	%a1,-(%sp)
 	movel	%a0,-(%sp)
-	jsr		copyin
+	jsr	copyin
 	addw	#12,%sp
 	movel	(%sp)+,%d1
 	rts
diff --git a/arch/m68k/ifpsp060/iskeleton.S b/arch/m68k/ifpsp060/iskeleton.S
index 5940a374a..3c310c237 100644
--- a/arch/m68k/ifpsp060/iskeleton.S
+++ b/arch/m68k/ifpsp060/iskeleton.S
@@ -36,13 +36,23 @@
 
 #include <linux/linkage.h>
 
-LOFF_ORIG_D0	= 0x20
+/*
+ * This has to match entry.S
+ */
+LOFF_ORIG_D0	= 0x24
+
+#define curptr a2
 
 #define SAVE_ALL				\
 	clrl	%sp@-;     /* stk_adj */	\
 	movel	%d0,%sp@-; /* orig d0 */	\
 	movel	%d0,%sp@-; /* d0 */		\
-	moveml	%d1-%d5/%a0-%a1,%sp@-
+	moveml	%d1-%d5/%a0-%a1/%curptr,%sp@-;
+
+#define GET_CURRENT(tmp) \
+	movel	%sp,tmp; \
+	andw	&-8192,tmp; \
+	movel	tmp,%curptr;
 
 |################################
 | (1) EXAMPLE CALL-OUTS 	#
@@ -85,6 +95,7 @@ Lmustsched:
 	SAVE_ALL
 	moveq	#-1,%d0
 	movel	%d0,%sp@(LOFF_ORIG_D0)    | indicate stack frame not for syscall
+	GET_CURRENT(%d0)
 	bral	SYMBOL_NAME(ret_from_exception)	| deliver signals, reschedule etc..
 
 |
diff --git a/arch/m68k/kernel/entry.S b/arch/m68k/kernel/entry.S
index 2cb35e67f..ef5ef46d6 100644
--- a/arch/m68k/kernel/entry.S
+++ b/arch/m68k/kernel/entry.S
@@ -33,12 +33,13 @@
  *	10(sp) - d5
  *	14(sp) - a0
  *	18(sp) - a1
- *	1C(sp) - d0
- *	20(sp) - orig_d0
- *	24(sp) - stack adjustment
- *	28(sp) - sr
- *	2A(sp) - pc
- *	2E(sp) - format & vector
+ *	1C(sp) - a2
+ *	20(sp) - d0
+ *	24(sp) - orig_d0
+ *	28(sp) - stack adjustment
+ *	2C(sp) - sr
+ *	2E(sp) - pc
+ *	32(sp) - format & vector
  */
 
 /*
@@ -47,6 +48,11 @@
  *               number 0 in the 'current_set' list.
  */
 
+/*
+ * 97/05/14 Andreas: Register %a2 is now set to the current task throughout
+ *		     the whole kernel.
+ */
+
 #include <linux/sys.h>
 #include <linux/config.h>
 #include <linux/linkage.h>
@@ -57,6 +63,8 @@
 .globl SYMBOL_NAME(kgdb_registers)
 #endif
 
+#define curptr a2
+
 LENOSYS = 38
 
 /*
@@ -80,15 +88,15 @@ LTASK_FLAGS	= 20
 #define	MAX_NOINT_IPL	0
 #endif /* machine compilation types */ 
 
-LD0		= 0x1C
-LORIG_D0	= 0x20
-LSR		= 0x28
-LFORMATVEC	= 0x2E
+LD0		= 0x20
+LORIG_D0	= 0x24
+LSR		= 0x2C
+LFORMATVEC	= 0x32
 
 /*
  * This defines the normal kernel pt-regs layout.
  *
- * regs are a2-a6 and d6-d7 preserved by C code
+ * regs a3-a6 and d6-d7 are preserved by C code
  * the kernel doesn't mess with usp unless it needs to
  */
 #ifndef CONFIG_KGDB
@@ -96,7 +104,7 @@ LFORMATVEC	= 0x2E
 	clrl	%sp@-;    /* stk_adj */	\
 	movel	%d0,%sp@-; /* orig d0 */	\
 	movel	%d0,%sp@-; /* d0 */	\
-	moveml	%d1-%d5/%a0-%a1,%sp@-
+	moveml	%d1-%d5/%a0-%a1/%curptr,%sp@-;
 #else
 /* Need to save the "missing" registers for kgdb...
  */
@@ -104,25 +112,30 @@ LFORMATVEC	= 0x2E
 	clrl	%sp@-;    /* stk_adj */				\
 	movel	%d0,%sp@-; /* orig d0 */			\
 	movel	%d0,%sp@-; /* d0 */				\
-	moveml	%d1-%d5/%a0-%a1,%sp@-;				\
+	moveml	%d1-%d5/%a0-%a1/%curptr,%sp@-;			\
 	moveml	%d6-%d7,SYMBOL_NAME(kgdb_registers)+GDBOFFA_D6;	\
-	moveml	%a2-%a6,SYMBOL_NAME(kgdb_registers)+GDBOFFA_A2
+	moveml	%a3-%a6,SYMBOL_NAME(kgdb_registers)+GDBOFFA_A3;
 #endif
 
 #define RESTORE_ALL			\
-	moveml	%sp@+,%a0-%a1/%d1-%d5;	\
+	moveml	%sp@+,%a0-%a1/%curptr/%d1-%d5;	\
 	movel	%sp@+,%d0;		\
 	addql	#4,%sp;	 /* orig d0 */	\
 	addl	%sp@+,%sp; /* stk adj */	\
 	rte
 
-#define SWITCH_STACK_SIZE (7*4+4)	/* includes return address */
+#define SWITCH_STACK_SIZE (6*4+4)	/* includes return address */
 
 #define SAVE_SWITCH_STACK \
-	moveml	%a2-%a6/%d6-%d7,%sp@-
+	moveml	%a3-%a6/%d6-%d7,%sp@-
 
 #define RESTORE_SWITCH_STACK \
-	moveml	%sp@+,%a2-%a6/%d6-%d7
+	moveml	%sp@+,%a3-%a6/%d6-%d7
+
+#define GET_CURRENT(tmp) \
+	movel	%sp,tmp; \
+	andw	&-8192,tmp; \
+	movel	tmp,%curptr;
 
 .globl SYMBOL_NAME(system_call), SYMBOL_NAME(buserr), SYMBOL_NAME(trap)
 .globl SYMBOL_NAME(resume), SYMBOL_NAME(ret_from_exception)
@@ -139,6 +152,7 @@ ENTRY(buserr)
 					| signifies that the stack frame
 					| is NOT for syscall
 
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(buserr_c)
 	addql	#4,%sp
@@ -150,6 +164,7 @@ ENTRY(trap)
 	movel	%d0,%sp@(LORIG_D0)	| a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	movel	%sp,%sp@- 		| stack frame pointer argument
 	bsrl	SYMBOL_NAME(trap_c)
 	addql	#4,%sp
@@ -190,6 +205,7 @@ ENTRY(system_call)
 	SAVE_ALL
 	movel	%d0,%d2
 
+	GET_CURRENT(%d0)
 	| save top of frame
 	pea	%sp@
 	jbsr	SYMBOL_NAME(set_esp0)
@@ -197,8 +213,7 @@ ENTRY(system_call)
 
 	cmpl	#NR_syscalls,%d2
 	jcc	badsys
-	movel	SYMBOL_NAME(current_set),%a0
-	btst	#5,%a0@(LTASK_FLAGS+3)	| PF_TRACESYS
+	btst	#5,%curptr@(LTASK_FLAGS+3)	| PF_TRACESYS
 	jne	do_trace
 	jbsr	@(SYMBOL_NAME(sys_call_table),%d2:l:4)@(0)
 	movel	%d0,%sp@(LD0)		| save the return value
@@ -208,24 +223,23 @@ SYMBOL_NAME_LABEL(ret_from_exception)
 	bnes	2f			| if so, skip resched, signals
 	tstl	SYMBOL_NAME(need_resched)
 	jne	SYMBOL_NAME(reschedule)
-	movel	SYMBOL_NAME(current_set),%a0
-	cmpl	#SYMBOL_NAME(task),%a0	| task[0] cannot have signals
+	cmpl	#SYMBOL_NAME(task),%curptr	| task[0] cannot have signals
 	jeq	2f
-	bclr	#5,%a0@(LTASK_FLAGS+1)	| check for delayed trace
+	bclr	#5,%curptr@(LTASK_FLAGS+1)	| check for delayed trace
 	jne	do_delayed_trace
 5:
-	tstl	%a0@(LTASK_STATE)	| state
+	tstl	%curptr@(LTASK_STATE)	| state
 	jne	SYMBOL_NAME(reschedule)
-	tstl	%a0@(LTASK_COUNTER)	| counter
+	tstl	%curptr@(LTASK_COUNTER)	| counter
 	jeq	SYMBOL_NAME(reschedule)
 
-	movel	%a0@(LTASK_BLOCKED),%d0
+	movel	%curptr@(LTASK_BLOCKED),%d0
 	movel	%d0,%d1			| save blocked in d1 for sig handling
 	notl	%d0
-	btst	#4,%a0@(LTASK_FLAGS+3)	| PF_PTRACED
+	btst	#4,%curptr@(LTASK_FLAGS+3)	| PF_PTRACED
 	jeq	1f
 	moveq	#-1,%d0			| let the debugger see all signals
-1:	andl	%a0@(LTASK_SIGNAL),%d0
+1:	andl	%curptr@(LTASK_SIGNAL),%d0
 	jne	Lsignal_return
 2:	RESTORE_ALL
 
@@ -248,7 +262,6 @@ do_delayed_trace:
 	jbsr	SYMBOL_NAME(send_sig)
 	addql	#8,%sp
 	addql	#4,%sp
-	movel	SYMBOL_NAME(current_set),%a0
 	jra	5b
 
 /*
@@ -260,6 +273,7 @@ SYMBOL_NAME_LABEL(inthandler)
 	movel	%d0,%sp@(LORIG_D0)	| a -1 in the ORIG_D0 field
 					| signifies that the stack frame
 					| is NOT for syscall
+	GET_CURRENT(%d0)
 	addql	#1,SYMBOL_NAME(local_irq_count)
 					|  put exception # in d0
 	bfextu %sp@(LFORMATVEC){#4,#10},%d0
@@ -270,28 +284,31 @@ SYMBOL_NAME_LABEL(inthandler)
 	addql	#8,%sp			|  pop parameters off stack
 
 SYMBOL_NAME_LABEL(ret_from_interrupt)
-	/* check if we need to do software interrupts */
-	movel	SYMBOL_NAME(local_irq_count),%d1
-	subql	#1,%d1
-	jne	4f
-#if 0
-		bfextu  %sp@(LSR){#5,#3},%d0    | Check for nested interrupt.
+	subql	#1,SYMBOL_NAME(local_irq_count)
+	jeq	1f
+2:	
+	RESTORE_ALL
+1:
+#if 1
+	bfextu  %sp@(LSR){#5,#3},%d0    | Check for nested interrupt.
 #if MAX_NOINT_IPL > 0
 	cmpiw	#MAX_NOINT_IPL,%d0
 #endif
-	jhi	4f
+	jhi	2b
 #endif
+	/* Let the rest run with interrupts allowed.  This is safe since
+	   the kernel never uses a non-standard ipl and this is the outer
+	   level interrupt. */
+	andw	#ALLOWINT,%sr
+
+	/* check if we need to do software interrupts */
+	
 	movel	SYMBOL_NAME(bh_active),%d0
 	andl	SYMBOL_NAME(bh_mask),%d0
-	jeq	3f
+	jeq	SYMBOL_NAME(ret_from_exception)
 
-	jbsr	SYMBOL_NAME(do_bottom_half)
-3:	
-	clrl	SYMBOL_NAME(local_irq_count)
-	jra	SYMBOL_NAME(ret_from_exception)
-4:
-	movel	%d1,SYMBOL_NAME(local_irq_count)
-	RESTORE_ALL
+	pea	SYMBOL_NAME(ret_from_exception)
+	jra	SYMBOL_NAME(do_bottom_half)
 
 
 /* Handler for uninitialized and spurious interrupts */
@@ -398,7 +415,7 @@ SYMBOL_NAME_LABEL(resume)
 3:
 
 	/* get pointer to tss struct (a1 contains new task) */
-	movel	%a1,SYMBOL_NAME(current_set)
+	movel	%a1,%curptr
 	addl	%d1,%a1
 
 	/* Skip address space switching if they are the same. */
@@ -419,7 +436,7 @@ SYMBOL_NAME_LABEL(resume)
 	movec	%cacr,%d0
 	oriw	#LFLUSH_I_AND_D,%d0
 	movec	%d0,%cacr
-	
+
 	/* switch the root pointer */
 	pmove	%a1@(LTSS_CRP),%crp
 #endif
diff --git a/arch/m68k/kernel/head.S b/arch/m68k/kernel/head.S
index c2b72aa2d..36a4072ac 100644
--- a/arch/m68k/kernel/head.S
+++ b/arch/m68k/kernel/head.S
@@ -223,17 +223,9 @@ ENTRY(_start)
 	movel	%d0,%a0@		/* save cache mode for page tables */
 
 /*
- * raise interrupt level with MASTER bit set, copy isp to msp (if not 68060)
+ * raise interrupt level
  */
-#ifdef FROM_PL9
-	movew	#0x3700,%sr
-	is_060(1f)
-	movec	%isp,%d0
-	movel	%d0,%sp
-1:
-#else
 	movew	#0x2700,%sr
-#endif
 
 /*
    If running on an Atari, determine the I/O base of the
@@ -896,8 +888,10 @@ Lcache68060:
 
 /*
  * Setup initial stack pointer
+ * We need to get current loaded up with our first task...
  */
-	lea	SYMBOL_NAME(init_user_stack)+PAGESIZE,%sp
+	lea	SYMBOL_NAME(init_task_union),%a2
+	lea	8192(%a2),%sp
 
 /* jump to the kernel start */
 	putr()
diff --git a/arch/m68k/kernel/ints.c b/arch/m68k/kernel/ints.c
index f49c38cd3..39e1fbce6 100644
--- a/arch/m68k/kernel/ints.c
+++ b/arch/m68k/kernel/ints.c
@@ -29,6 +29,7 @@
 #include <linux/sched.h>
 #include <linux/kernel_stat.h>
 #include <linux/errno.h>
+#include <linux/init.h>
 
 #include <asm/system.h>
 #include <asm/irq.h>
@@ -79,7 +80,7 @@ void (*mach_free_irq) (unsigned int, void *) = dummy_free_irq;
  * the IRQ handling routines.
  */
 
-void init_IRQ(void)
+__initfunc(void init_IRQ(void))
 {
 	int i;
 
diff --git a/arch/m68k/kernel/m68k_ksyms.c b/arch/m68k/kernel/m68k_ksyms.c
index 3138d99df..f6bb0689b 100644
--- a/arch/m68k/kernel/m68k_ksyms.c
+++ b/arch/m68k/kernel/m68k_ksyms.c
@@ -6,6 +6,8 @@
 #include <linux/mm.h>
 #include <linux/user.h>
 #include <linux/elfcore.h>
+#include <linux/in6.h>
+#include <linux/interrupt.h>
 
 #include <asm/setup.h>
 #include <asm/machdep.h>
@@ -14,6 +16,7 @@
 #include <asm/semaphore.h>
 #include <asm/checksum.h>
 #include <asm/hardirq.h>
+#include <asm/softirq.h>
 
 asmlinkage long long __ashrdi3 (long long, int);
 extern char m68k_debug_device[];
@@ -40,6 +43,7 @@ EXPORT_SYMBOL(strnlen);
 EXPORT_SYMBOL(strrchr);
 EXPORT_SYMBOL(strstr);
 EXPORT_SYMBOL(local_irq_count);
+EXPORT_SYMBOL(__m68k_bh_counter);
 
 /* Networking helper routines. */
 EXPORT_SYMBOL(csum_partial_copy);
diff --git a/arch/m68k/kernel/process.c b/arch/m68k/kernel/process.c
index 5422831a8..ad0662f22 100644
--- a/arch/m68k/kernel/process.c
+++ b/arch/m68k/kernel/process.c
@@ -10,6 +10,7 @@
  * This file handles the architecture-dependent parts of process handling..
  */
 
+#include <linux/config.h>
 #include <linux/errno.h>
 #include <linux/sched.h>
 #include <linux/kernel.h>
@@ -29,6 +30,23 @@
 #include <asm/traps.h>
 #include <asm/machdep.h>
 #include <asm/setup.h>
+#include <asm/pgtable.h>
+
+/*
+ * Initial task structure. Make this a per-architecture thing,
+ * because different architectures tend to have different
+ * alignment requirements and potentially different initial
+ * setup.
+ */
+static struct vm_area_struct init_mmap = INIT_MMAP;
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS;
+struct mm_struct init_mm = INIT_MM;
+
+union task_union init_task_union
+	__attribute__((section("init_task"), aligned(2*PAGE_SIZE)))
+	= { task: INIT_TASK };
 
 asmlinkage void ret_from_exception(void);
 
@@ -46,8 +64,17 @@ asmlinkage int sys_idle(void)
 	/* endless idle loop with no priority at all */
 	current->priority = -100;
 	current->counter = -100;
-	for (;;)
+	for (;;){
+		if (!need_resched)
+#if defined(CONFIG_ATARI) && !defined(CONFIG_AMIGA) && !defined(CONFIG_MAC)
+			/* block out HSYNC on the atari (falcon) */
+			__asm__("stop #0x2200" : : : "cc");
+#else /* portable version */
+			__asm__("stop #0x2000" : : : "cc");
+#endif /* machine compilation types */ 
+		run_task_queue(&tq_scheduler);
 		schedule();
+	}
 	ret = 0;
 out:
 	unlock_kernel();
@@ -76,8 +103,8 @@ void show_regs(struct pt_regs * regs)
 	printk("\n");
 	printk("Format %02x  Vector: %04x  PC: %08lx  Status: %04x\n",
 	       regs->format, regs->vector, regs->pc, regs->sr);
-	printk("ORIG_D0: %08lx  D0: %08lx  A1: %08lx\n",
-	       regs->orig_d0, regs->d0, regs->a1);
+	printk("ORIG_D0: %08lx  D0: %08lx  A2: %08lx  A1: %08lx\n",
+	       regs->orig_d0, regs->d0, regs->a2, regs->a1);
 	printk("A0: %08lx  D5: %08lx  D4: %08lx\n",
 	       regs->a0, regs->d5, regs->d4);
 	printk("D3: %08lx  D2: %08lx  D1: %08lx\n",
@@ -144,8 +171,8 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 	struct switch_stack * childstack, *stack;
 	unsigned long stack_offset, *retp;
 
-	stack_offset = PAGE_SIZE - sizeof(struct pt_regs);
-	childregs = (struct pt_regs *) (p->kernel_stack_page + stack_offset);
+	stack_offset = 2*PAGE_SIZE - sizeof(struct pt_regs);
+	childregs = (struct pt_regs *) ((unsigned long) p + stack_offset);
 
 	*childregs = *regs;
 	childregs->d0 = 0;
@@ -231,7 +258,7 @@ void dump_thread(struct pt_regs * regs, struct user * dump)
 	dump->regs.d7 = sw->d7;
 	dump->regs.a0 = regs->a0;
 	dump->regs.a1 = regs->a1;
-	dump->regs.a2 = sw->a2;
+	dump->regs.a2 = regs->a2;
 	dump->regs.a3 = sw->a3;
 	dump->regs.a4 = sw->a4;
 	dump->regs.a5 = sw->a5;
diff --git a/arch/m68k/kernel/ptrace.c b/arch/m68k/kernel/ptrace.c
index 70e341b31..be4149cbb 100644
--- a/arch/m68k/kernel/ptrace.c
+++ b/arch/m68k/kernel/ptrace.c
@@ -47,23 +47,11 @@
 static int regoff[] = {
 	PT_REG(d1), PT_REG(d2), PT_REG(d3), PT_REG(d4),
 	PT_REG(d5), SW_REG(d6), SW_REG(d7), PT_REG(a0),
-	PT_REG(a1), SW_REG(a2), SW_REG(a3), SW_REG(a4),
+	PT_REG(a1), PT_REG(a2), SW_REG(a3), SW_REG(a4),
 	SW_REG(a5), SW_REG(a6), PT_REG(d0), -1,
 	PT_REG(orig_d0), PT_REG(sr), PT_REG(pc),
 };
 
-/* change a pid into a task struct. */
-static inline struct task_struct * get_task(int pid)
-{
-	int i;
-
-	for (i = 1; i < NR_TASKS; i++) {
-		if (task[i] != NULL && (task[i]->pid == pid))
-			return task[i];
-	}
-	return NULL;
-}
-
 /*
  * Get contents of register REGNO in task TASK.
  */
@@ -116,7 +104,7 @@ static unsigned long get_long(struct task_struct * tsk,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (pgd_none(*pgdir)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -126,7 +114,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir,addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -137,7 +125,7 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
@@ -168,7 +156,7 @@ static void put_long(struct task_struct * tsk, struct vm_area_struct * vma, unsi
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (!pgd_present(*pgdir)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -178,7 +166,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir,addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -189,12 +177,12 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
 	if (!pte_write(*pgtable)) {
-		do_wp_page(tsk, vma, addr, 2);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 /* this is a hack for non-kernel-mapped video buffers and similar */
@@ -340,7 +328,7 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 	if (pid == 1)		/* you may not mess with init */
 		goto out;
 	ret = -ESRCH;
-	if (!(child = get_task(pid)))
+	if (!(child = find_task_by_pid(pid)))
 		goto out;
 	ret = -EPERM;
 	if (request == PTRACE_ATTACH) {
diff --git a/arch/m68k/kernel/setup.c b/arch/m68k/kernel/setup.c
index dfd91d0d4..def50a747 100644
--- a/arch/m68k/kernel/setup.c
+++ b/arch/m68k/kernel/setup.c
@@ -18,6 +18,7 @@
 #include <linux/genhd.h>
 #include <linux/errno.h>
 #include <linux/string.h>
+#include <linux/init.h>
 
 #include <asm/bootinfo.h>
 #include <asm/setup.h>
@@ -55,13 +56,13 @@ static struct mem_info m68k_ramdisk = { 0, 0 };
 static char m68k_command_line[CL_SIZE];
 char saved_command_line[CL_SIZE];
 
-void (*mach_sched_init) (void (*handler)(int, void *, struct pt_regs *));
+void (*mach_sched_init) (void (*handler)(int, void *, struct pt_regs *)) __initdata;
 /* machine dependent keyboard functions */
-int (*mach_keyb_init) (void);
+int (*mach_keyb_init) (void) __initdata;
 int (*mach_kbdrate) (struct kbd_repeat *) = NULL;
 void (*mach_kbd_leds) (unsigned int) = NULL;
 /* machine dependent irq functions */
-void (*mach_init_IRQ) (void);
+void (*mach_init_IRQ) (void) __initdata;
 void (*(*mach_default_handler)[]) (int, void *, struct pt_regs *) = NULL;
 void (*mach_get_model) (char *model) = NULL;
 int (*mach_get_hardware_list) (char *buffer) = NULL;
@@ -73,12 +74,12 @@ void (*mach_gettod) (int*, int*, int*, int*, int*, int*);
 int (*mach_hwclk) (int, struct hwclk_time*) = NULL;
 int (*mach_set_clock_mmss) (unsigned long) = NULL;
 void (*mach_reset)( void );
-struct fb_info *(*mach_fb_init)(long *);
+struct fb_info *(*mach_fb_init)(long *) __initdata;
 long mach_max_dma_address = 0x00ffffff; /* default set to the lower 16MB */
-void (*mach_video_setup) (char *, int *);
+void (*mach_video_setup) (char *, int *) __initdata;
 #ifdef CONFIG_BLK_DEV_FD
-int (*mach_floppy_init) (void) = NULL;
-void (*mach_floppy_setup) (char *, int *) = NULL;
+int (*mach_floppy_init) (void) __initdata = NULL;
+void (*mach_floppy_setup) (char *, int *) __initdata = NULL;
 void (*mach_floppy_eject) (void) = NULL;
 #endif
 
@@ -94,7 +95,7 @@ extern void config_apollo(void);
 #define MASK_256K 0xfffc0000
 
 
-static void m68k_parse_bootinfo(const struct bi_record *record)
+__initfunc(static void m68k_parse_bootinfo(const struct bi_record *record))
 {
     while (record->tag != BI_LAST) {
 	int unknown = 0;
@@ -141,8 +142,8 @@ static void m68k_parse_bootinfo(const struct bi_record *record)
     }
 }
 
-void setup_arch(char **cmdline_p, unsigned long * memory_start_p,
-		unsigned long * memory_end_p)
+ __initfunc(void setup_arch(char **cmdline_p, unsigned long * memory_start_p,
+		unsigned long * memory_end_p))
 {
 	unsigned long memory_start, memory_end;
 	extern int _etext, _edata, _end;
@@ -337,7 +338,7 @@ int get_hardware_list(char *buffer)
 }
 
 #ifdef CONFIG_BLK_DEV_FD
-int floppy_init(void)
+__initfunc(int floppy_init(void))
 {
 	if (mach_floppy_init)
 		return mach_floppy_init();
@@ -345,7 +346,7 @@ int floppy_init(void)
 		return 0;
 }
 
-void floppy_setup(char *str, int *ints)
+__initfunc(void floppy_setup(char *str, int *ints))
 {
 	if (mach_floppy_setup)
 		mach_floppy_setup (str, ints);
@@ -358,7 +359,7 @@ void floppy_eject(void)
 }
 #endif
 
-unsigned long arch_kbd_init(void)
+__initfunc(unsigned long arch_kbd_init(void))
 {
 	return mach_keyb_init();
 }
@@ -372,7 +373,7 @@ void arch_gettod(int *year, int *mon, int *day, int *hour,
 		*year = *mon = *day = *hour = *min = *sec = 0;
 }
 
-void video_setup (char *options, int *ints)
+__initfunc(void video_setup (char *options, int *ints))
 {
 	if (mach_video_setup)
 		mach_video_setup (options, ints);
diff --git a/arch/m68k/kernel/signal.c b/arch/m68k/kernel/signal.c
index c60e82b0b..52c13445d 100644
--- a/arch/m68k/kernel/signal.c
+++ b/arch/m68k/kernel/signal.c
@@ -42,7 +42,9 @@
 
 #define _BLOCKABLE (~(_S(SIGKILL) | _S(SIGSTOP)))
 
-asmlinkage int sys_waitpid(pid_t pid,unsigned long * stat_addr, int options);
+asmlinkage int sys_wait4(pid_t pid, unsigned long *stat_addr,
+			 int options, unsigned long *ru);
+
 asmlinkage int do_signal(unsigned long oldmask, struct pt_regs *regs);
 
 const int frame_extra_sizes[16] = {
@@ -466,7 +468,7 @@ asmlinkage int do_signal(unsigned long oldmask, struct pt_regs *regs)
 			if (signr != SIGCHLD)
 				continue;
 			/* check for SIGCHLD: it's special */
-			while (sys_waitpid(-1,NULL,WNOHANG) > 0)
+			while (sys_wait4(-1,NULL,WNOHANG, NULL) > 0)
 				/* nothing */;
 			continue;
 		}
diff --git a/arch/m68k/kernel/sys_m68k.c b/arch/m68k/kernel/sys_m68k.c
index 5acfd1cbd..e29509cac 100644
--- a/arch/m68k/kernel/sys_m68k.c
+++ b/arch/m68k/kernel/sys_m68k.c
@@ -532,12 +532,15 @@ sys_cacheflush (unsigned long addr, int scope, int cache, unsigned long len)
 		 */
 		vma = find_vma (current->mm, addr);
 		ret = -EINVAL;
+		/* Check for overflow.  */
+		if (addr + len < addr)
+			goto out;
 		if (vma == NULL || addr < vma->vm_start || addr + len > vma->vm_end)
 			goto out;
 	}
 
 	if (CPU_IS_020_OR_030) {
-		if (scope == FLUSH_SCOPE_LINE) {
+		if (scope == FLUSH_SCOPE_LINE && len < 256) {
 			unsigned long cacr;
 			__asm__ ("movec %%cacr, %0" : "=r" (cacr));
 			if (cache & FLUSH_CACHE_INSN)
diff --git a/arch/m68k/kernel/time.c b/arch/m68k/kernel/time.c
index dbff49276..7b3acdfa6 100644
--- a/arch/m68k/kernel/time.c
+++ b/arch/m68k/kernel/time.c
@@ -27,6 +27,24 @@ static inline int set_rtc_mmss(unsigned long nowtime)
   return -1;
 }
 
+static inline void do_profile (unsigned long pc)
+{
+	if (prof_buffer && current->pid) {
+		extern int _stext;
+		pc -= (unsigned long) &_stext;
+		pc >>= prof_shift;
+		if (pc < prof_len)
+			++prof_buffer[pc];
+		else
+		/*
+		 * Dont ignore out-of-bounds PC values silently,
+		 * put them into the last histogram slot, so if
+		 * present, they will show up as a sharp peak.
+		 */
+			++prof_buffer[prof_len-1];
+	}
+}
+
 /*
  * timer_interrupt() needs to keep up the real-time clock,
  * as well as call the "do_timer()" routine every clocktick
@@ -38,6 +56,9 @@ static void timer_interrupt(int irq, void *dummy, struct pt_regs * regs)
 
 	do_timer(regs);
 
+	if (!user_mode(regs))
+		do_profile(regs->pc);
+
 	/*
 	 * If we have an externally synchronized Linux clock, then update
 	 * CMOS clock accordingly every ~11 minutes. Set_rtc_mmss() has to be
diff --git a/arch/m68k/kernel/traps.c b/arch/m68k/kernel/traps.c
index 1cc547907..25be40007 100644
--- a/arch/m68k/kernel/traps.c
+++ b/arch/m68k/kernel/traps.c
@@ -27,6 +27,7 @@
 #include <linux/user.h>
 #include <linux/string.h>
 #include <linux/linkage.h>
+#include <linux/init.h>
 
 #include <asm/setup.h>
 #include <asm/fpu.h>
@@ -63,7 +64,7 @@ asm(".text\n"
     __ALIGN_STR "\n"
     SYMBOL_NAME_STR(nmihandler) ": rte");
 
-void trap_init (void)
+__initfunc(void trap_init (void))
 {
 	int i;
 
@@ -932,16 +933,15 @@ void die_if_kernel (char *str, struct pt_regs *fp, int nr)
 #endif
 	console_verbose();
 	printk("%s: %08x\n",str,nr);
-	printk("PC: [<%08lx>]\nSR: %04x  SP: %p\n", fp->pc, fp->sr, fp);
+	printk("PC: [<%08lx>]\nSR: %04x  SP: %p  a2: %08lx\n",
+	       fp->pc, fp->sr, fp, fp->a2);
 	printk("d0: %08lx    d1: %08lx    d2: %08lx    d3: %08lx\n",
 	       fp->d0, fp->d1, fp->d2, fp->d3);
 	printk("d4: %08lx    d5: %08lx    a0: %08lx    a1: %08lx\n",
 	       fp->d4, fp->d5, fp->a0, fp->a1);
 
-	if (STACK_MAGIC != *(unsigned long *)current->kernel_stack_page)
-		printk("Corrupted stack page\n");
 	printk("Process %s (pid: %d, stackpage=%08lx)\n",
-		current->comm, current->pid, current->kernel_stack_page);
+		current->comm, current->pid, PAGE_SIZE+(unsigned long)current);
 #ifdef CONFIG_KGDB
 	}
 #endif
diff --git a/arch/m68k/lib/semaphore.S b/arch/m68k/lib/semaphore.S
index 76ffc3cc5..8e4141149 100644
--- a/arch/m68k/lib/semaphore.S
+++ b/arch/m68k/lib/semaphore.S
@@ -19,8 +19,7 @@ ENTRY(__down_failed)
 	movel %a1,-(%sp)
 	jbsr SYMBOL_NAME(__down)
 	movel (%sp)+,%a1
-	movel (%sp)+,%d0
-	movel (%sp)+,%d1
+	moveml (%sp)+,%a0/%d0/%d1
 	rts
 
 ENTRY(__down_failed_interruptible)
@@ -30,6 +29,7 @@ ENTRY(__down_failed_interruptible)
 	jbsr SYMBOL_NAME(__down_interruptible)
 	movel (%sp)+,%a1
 	movel (%sp)+,%d1
+	movel (%sp)+,%a0
 	rts
 
 ENTRY(__up_wakeup)
@@ -37,6 +37,5 @@ ENTRY(__up_wakeup)
 	movel %a1,-(%sp)
 	jbsr SYMBOL_NAME(__up)
 	movel (%sp)+,%a1
-	movel (%sp)+,%d0
-	movel (%sp)+,%d1
+	moveml (%sp)+,%a0/%d0/%d1
 	rts
diff --git a/arch/m68k/mm/fault.c b/arch/m68k/mm/fault.c
index 71b46e2d9..193c6baee 100644
--- a/arch/m68k/mm/fault.c
+++ b/arch/m68k/mm/fault.c
@@ -32,14 +32,10 @@ extern const int frame_extra_sizes[]; /* in m68k/kernel/signal.c */
 asmlinkage int do_page_fault(struct pt_regs *regs, unsigned long address,
 			      unsigned long error_code)
 {
-	void (*handler)(struct task_struct *,
-			struct vm_area_struct *,
-			unsigned long,
-			int);
 	struct task_struct *tsk = current;
 	struct mm_struct *mm = tsk->mm;
 	struct vm_area_struct * vma;
-	unsigned long fixup, fault_pc;
+	unsigned long fixup;
 	int write;
 
 #ifdef DEBUG
@@ -73,10 +69,8 @@ asmlinkage int do_page_fault(struct pt_regs *regs, unsigned long address,
  */
 good_area:
 	write = 0;
-	handler = do_no_page;
 	switch (error_code & 3) {
 		default:	/* 3: write, present */
-			handler = do_wp_page;
 			/* fall through */
 		case 2:		/* write, not present */
 			if (!(vma->vm_flags & VM_WRITE))
@@ -89,7 +83,7 @@ good_area:
 			if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
 				goto bad_area;
 	}
-	handler(tsk, vma, address, write);
+	handle_mm_fault(current, vma, address, write);
 	up(&mm->mmap_sem);
 
 	/* There seems to be a missing invalidate somewhere in do_no_page.
@@ -108,10 +102,10 @@ bad_area:
 	up(&mm->mmap_sem);
 
 	/* Are we prepared to handle this fault?  */
-	fault_pc = regs->pc;
-	if ((fixup = search_exception_table(fault_pc)) != 0) {
+	if ((fixup = search_exception_table(regs->pc)) != 0) {
 		struct pt_regs *tregs;
-		printk(KERN_DEBUG "Exception at [<%lx>] (%lx)\n", fault_pc, fixup);
+		printk(KERN_DEBUG "%s: Exception at [<%lx>] (%lx)\n",
+		       current->comm, regs->pc, fixup);
 		/* Create a new four word stack frame, discarding the old
 		   one.  */
 		regs->stkadj = frame_extra_sizes[regs->format];
diff --git a/arch/m68k/mm/init.c b/arch/m68k/mm/init.c
index 01cd315dd..b46037f80 100644
--- a/arch/m68k/mm/init.c
+++ b/arch/m68k/mm/init.c
@@ -12,6 +12,7 @@
 #include <linux/kernel.h>
 #include <linux/string.h>
 #include <linux/types.h>
+#include <linux/init.h>
 #ifdef CONFIG_BLK_DEV_RAM
 #include <linux/blk.h>
 #endif
@@ -116,9 +117,8 @@ pte_t *kernel_page_table (unsigned long *memavailp)
 	return ptablep;
 }
 
-static unsigned long map_chunk (unsigned long addr,
-				unsigned long size,
-				unsigned long *memavailp)
+__initfunc(static unsigned long
+map_chunk (unsigned long addr, unsigned long size, unsigned long *memavailp))
 {
 #define ONEMEG	(1024*1024)
 #define L3TREESIZE (256*1024)
@@ -283,6 +283,11 @@ static unsigned long map_chunk (unsigned long addr,
 
 extern unsigned long free_area_init(unsigned long, unsigned long);
 
+/* References to section boundaries */
+
+extern char _text, _etext, _edata, __bss_start, _end;
+extern char __init_begin, __init_end;
+
 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
 /*
@@ -291,7 +296,7 @@ extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
  * The parameters are pointers to where to stick the starting and ending
  * addresses  of available kernel virtual memory.
  */
-unsigned long paging_init(unsigned long start_mem, unsigned long end_mem)
+__initfunc(unsigned long paging_init(unsigned long start_mem, unsigned long end_mem))
 {
 	int chunk;
 	unsigned long mem_avail = 0;
@@ -395,12 +400,12 @@ unsigned long paging_init(unsigned long start_mem, unsigned long end_mem)
 	return PAGE_ALIGN(free_area_init (start_mem, end_mem));
 }
 
-void mem_init(unsigned long start_mem, unsigned long end_mem)
+__initfunc(void mem_init(unsigned long start_mem, unsigned long end_mem))
 {
 	int codepages = 0;
 	int datapages = 0;
+	int initpages = 0;
 	unsigned long tmp;
-	extern int _etext;
 
 	end_mem &= PAGE_MASK;
 	high_memory = (void *) end_mem;
@@ -448,8 +453,15 @@ void mem_init(unsigned long start_mem, unsigned long end_mem)
 		if (VTOP (tmp) >= mach_max_dma_address)
 			clear_bit(PG_DMA, &mem_map[MAP_NR(tmp)].flags);
 		if (PageReserved(mem_map+MAP_NR(tmp))) {
-			if (tmp < (unsigned long)&_etext)
-				codepages++;
+			if (tmp >= (unsigned long)&_text
+			    && tmp < (unsigned long)&_edata) {
+				if (tmp < (unsigned long) &_etext)
+					codepages++;
+				else
+					datapages++;
+			} else if (tmp >= (unsigned long) &__init_begin
+				   && tmp < (unsigned long) &__init_end)
+				initpages++;
 			else
 				datapages++;
 			continue;
@@ -461,16 +473,24 @@ void mem_init(unsigned long start_mem, unsigned long end_mem)
 #endif
 			free_page(tmp);
 	}
-	printk("Memory: %luk/%luk available (%dk kernel code, %dk data)\n",
+	printk("Memory: %luk/%luk available (%dk kernel code, %dk data, %dk init)\n",
 	       (unsigned long) nr_free_pages << (PAGE_SHIFT-10),
 	       max_mapnr << (PAGE_SHIFT-10),
 	       codepages << (PAGE_SHIFT-10),
-	       datapages << (PAGE_SHIFT-10));
+	       datapages << (PAGE_SHIFT-10),
+	       initpages << (PAGE_SHIFT-10));
 }
 
 void free_initmem(void)
 {
-	/* To be written */
+	unsigned long addr;
+
+	addr = (unsigned long)&__init_begin;
+	for (; addr < (unsigned long)&__init_end; addr += PAGE_SIZE) {
+		mem_map[MAP_NR(addr)].flags &= ~(1 << PG_reserved);
+		atomic_set(&mem_map[MAP_NR(addr)].count, 1);
+		free_page(addr);
+	}
 }
 
 void si_meminfo(struct sysinfo *val)
diff --git a/arch/m68k/mm/memory.c b/arch/m68k/mm/memory.c
index 55d5e98f3..77eb2dbfe 100644
--- a/arch/m68k/mm/memory.c
+++ b/arch/m68k/mm/memory.c
@@ -280,6 +280,8 @@ static unsigned long transp_transl_matches( unsigned long regval,
     return( (vaddr & mask) == (base & mask) );
 }
 
+static unsigned long mm_vtop_fallback (unsigned long);
+
 /*
  * The following two routines map from a physical address to a kernel
  * virtual address and vice versa.
@@ -301,7 +303,13 @@ unsigned long mm_vtop (unsigned long vaddr)
 			offset += m68k_memory[i].size;
 		i++;
 	}while (i < m68k_num_memory);
+	return mm_vtop_fallback(vaddr);
+}
 
+/* Separate function to make the common case faster (needs to save less
+   registers) */
+static unsigned long mm_vtop_fallback (unsigned long vaddr)
+{
 	/* not in one of the memory chunks; test for applying transparent
 	 * translation */
 
diff --git a/arch/m68k/vmlinux.lds b/arch/m68k/vmlinux.lds
new file mode 100644
index 000000000..cc5b1e20e
--- /dev/null
+++ b/arch/m68k/vmlinux.lds
@@ -0,0 +1,60 @@
+/* ld script to make m68k Linux kernel */
+OUTPUT_FORMAT("elf32-m68k", "elf32-m68k", "elf32-m68k")
+OUTPUT_ARCH(m68k)
+ENTRY(_start)
+SECTIONS
+{
+  . = 0x1000;
+  _text = .;			/* Text and read-only data */
+  .text : {
+	*(.text)
+	*(.fixup)
+	*(.text.lock)		/* out-of-line lock text */
+	*(.gnu.warning)
+	} = 0x4e75
+  .rodata : { *(.rodata) }
+  .kstrtab : { *(.kstrtab) }
+
+  . = ALIGN(16);		/* Exception table */
+  __start___ex_table = .;
+  __ex_table : { *(__ex_table) }
+  __stop___ex_table = .;
+
+  __start___ksymtab = .;	/* Kernel symbol table */
+  __ksymtab : { *(__ksymtab) }
+  __stop___ksymtab = .;
+
+  _etext = .;			/* End of text section */
+
+  .data : {			/* Data */
+	*(.data)
+	CONSTRUCTORS
+	}
+
+  . = ALIGN(8192);
+  init_task : { *(init_task) }	/* The initial task and kernel stack */
+
+  _edata = .;			/* End of data section */
+
+  . = ALIGN(4096);		/* Init code and data */
+  __init_begin = .;
+  .text.init : { *(.text.init) }
+  .data.init : { *(.data.init) }
+  . = ALIGN(4096);
+  __init_end = .;
+
+  __bss_start = .;		/* BSS */
+  .bss : {
+	*(.bss)
+	}
+  _end = . ;
+
+  /* Stabs debugging sections.  */
+  .stab 0 : { *(.stab) }
+  .stabstr 0 : { *(.stabstr) }
+  .stab.excl 0 : { *(.stab.excl) }
+  .stab.exclstr 0 : { *(.stab.exclstr) }
+  .stab.index 0 : { *(.stab.index) }
+  .stab.indexstr 0 : { *(.stab.indexstr) }
+  .comment 0 : { *(.comment) }
+}
diff --git a/arch/mips/Makefile b/arch/mips/Makefile
index 5ea699450..1aa371ae1 100644
--- a/arch/mips/Makefile
+++ b/arch/mips/Makefile
@@ -156,9 +156,9 @@ endif
 # The pipe options is bad for my low-mem machine
 # Uncomment this if you want this.
 #
-#CFLAGS		+= -pipe
+CFLAGS		+= -pipe
 
-HEAD := arch/mips/kernel/head.o
+HEAD := arch/mips/kernel/head.o arch/mips/kernel/init_task.o
 
 SUBDIRS := $(SUBDIRS) $(addprefix arch/mips/, kernel mm lib tools)
 CORE_FILES := arch/mips/kernel/kernel.o arch/mips/mm/mm.o $(CORE_FILES)
diff --git a/arch/mips/boot/Makefile b/arch/mips/boot/Makefile
index fd445c865..dd834efd9 100644
--- a/arch/mips/boot/Makefile
+++ b/arch/mips/boot/Makefile
@@ -34,7 +34,11 @@ mkboot: mkboot.c
 	$(HOSTCC) -o $@ $^
 
 zdisk:	zImage
-	mcopy -o zImage a:vmlinux
+	if [ -f /etc/remote-mcopy ]; then \
+		ssh rio mcopy -o - a:vmlinux <zImage; \
+	else \
+		mcopy -o zImage a:vmlinux; \
+	fi
 
 dep:
 	$(CPP) -M *.[cS] > .depend
diff --git a/arch/mips/boot/mkboot.c b/arch/mips/boot/mkboot.c
index 48f27113e..35612d248 100644
--- a/arch/mips/boot/mkboot.c
+++ b/arch/mips/boot/mkboot.c
@@ -355,7 +355,7 @@ main(argc, argv)
 	char *infile, *outfile;
 	struct stat ifstat;
 	off_t ifsize;
-	void *image;
+	char *image;
 	int ifd, ofd, i, symtabix, strtabix;
 	Elf32_Ehdr eh;
 	Elf32_Phdr *ph;
@@ -421,7 +421,7 @@ main(argc, argv)
 	 * we're reading might have different type sizes, byteorder
 	 * or alignment than the host.
 	 */
-	memcpy(eh.e_ident, image, sizeof(eh.e_ident));
+	memcpy(eh.e_ident, (void *)image, sizeof(eh.e_ident));
 	if(memcmp(eh.e_ident, ELFMAG, SELFMAG)) {
 		fprintf(stderr, "Input file isn't a ELF file\n");
 		exit(1);
@@ -476,7 +476,7 @@ main(argc, argv)
 		exit(1);
 	}
 	for(i = 0;i < eh.e_phnum; i++)
-		get_elfph(image + eh.e_phoff + i * 32, ph + i);
+		get_elfph((void *)(image + eh.e_phoff + i * 32), ph + i);
 
 	/*
 	 * ... and then the section headers.
@@ -487,7 +487,7 @@ main(argc, argv)
 		exit(1);
 	}
 	for(i = 0;i < eh.e_shnum; i++)
-		get_elfsh(image + eh.e_shoff + (i * 40), sh + i);
+		get_elfsh((void *)(image + eh.e_shoff + (i * 40)), sh + i);
 
 	/*
 	 * Find the symboltable and the stringtable in the file.
diff --git a/arch/mips/config.in b/arch/mips/config.in
index a0e655ea1..14e9dec3f 100644
--- a/arch/mips/config.in
+++ b/arch/mips/config.in
@@ -178,23 +178,27 @@ fi
 source fs/Config.in
 
 if [ "$CONFIG_SGI" != "y" ]; then
-	source drivers/char/Config.in
+  source drivers/char/Config.in
 
-	mainmenu_option next_comment
-	comment 'Sound'
+  mainmenu_option next_comment
+  comment 'Sound'
 
-	tristate 'Sound card support' CONFIG_SOUND
-	if [ "$CONFIG_SOUND" != "n" ]; then
-	  source drivers/sound/Config.in
-	fi
-	endmenu
+  tristate 'Sound card support' CONFIG_SOUND
+  if [ "$CONFIG_SOUND" != "n" ]; then
+    source drivers/sound/Config.in
+  fi
+  endmenu
 else
-	comment 'SGI Character Devices'
-	tristate 'PS/2 mouse (aka "auxiliary device") support' CONFIG_PSMOUSE
-	bool 'SGI Zilog85C30 serial support' CONFIG_SGI_SERIAL
-	if [ "$CONFIG_SGI_SERIAL" != "n" ]; then
-		define_bool CONFIG_SERIAL y
-	fi
+  comment 'SGI Character Devices'
+  bool 'Virtual terminal' CONFIG_VT
+  if [ "$CONFIG_VT" = "y" ]; then
+    bool 'Console on virtual terminal' CONFIG_VT_CONSOLE
+  fi
+  tristate 'PS/2 mouse (aka "auxiliary device") support' CONFIG_PSMOUSE
+  bool 'SGI Zilog85C30 serial support' CONFIG_SGI_SERIAL
+  if [ "$CONFIG_SGI_SERIAL" != "n" ]; then
+    define_bool CONFIG_SERIAL y
+  fi
 fi
 
 mainmenu_option next_comment
diff --git a/arch/mips/defconfig b/arch/mips/defconfig
index dc28a8cd8..33f54a301 100644
--- a/arch/mips/defconfig
+++ b/arch/mips/defconfig
@@ -22,6 +22,7 @@ CONFIG_PCI=y
 #
 # CONFIG_CPU_R3000 is not set
 # CONFIG_CPU_R6000 is not set
+# CONFIG_CPU_R4300 is not set
 CONFIG_CPU_R4X00=y
 # CONFIG_CPU_R5000 is not set
 # CONFIG_CPU_R8000 is not set
diff --git a/arch/mips/deskstation/io.c b/arch/mips/deskstation/io.c
new file mode 100644
index 000000000..cd33e2eb6
--- /dev/null
+++ b/arch/mips/deskstation/io.c
@@ -0,0 +1,68 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Low level I/O functions for Jazz family machine.
+ *
+ * FIXME: This implementation fits the Tyne.  How does the EISA rPC44 handle
+ * the eight high address bits?
+ */
+#include <linux/string.h>
+#include <asm/mipsconfig.h>
+#include <asm/addrspace.h>
+#include <asm/sni.h>
+
+/*
+ * isa_slot_offset is the address where E(ISA) busaddress 0 is is mapped
+ * for the processor.
+ */
+extern unsigned long isa_slot_offset;
+
+static unsigned char deskstation_readb(unsigned long addr)
+{
+	return *(volatile unsigned char *) (isa_slot_offset + addr);
+}
+
+static unsigned short deskstation_readw(unsigned long addr)
+{
+	return *(volatile unsigned short *) (isa_slot_offset + addr);
+}
+
+static unsigned int deskstation_readl(unsigned long addr)
+{
+	return *(volatile unsigned int *) (isa_slot_offset + addr);
+}
+
+static void deskstation_writeb(unsigned char val, unsigned long addr)
+{
+	*(volatile unsigned char *) (isa_slot_offset + addr) = val;
+}
+
+static void deskstation_writew(unsigned short val, unsigned long addr)
+{
+	*(volatile unsigned char *) (isa_slot_offset + addr) = val;
+}
+
+static void deskstation_writel(unsigned int val, unsigned long addr)
+{
+	*(volatile unsigned char *) (isa_slot_offset + addr) = val;
+}
+
+static void deskstation_memset_io(unsigned long addr, int val, unsigned long len)
+{
+	addr += isa_slot_offset;
+	memset((void *)addr, val, len);
+}
+
+static void deskstation_memcpy_fromio(unsigned long to, unsigned long from, unsigned long len)
+{
+	from += isa_slot_offset;
+	memcpy((void *)to, (void *)from, len);
+}
+
+static void deskstation_memcpy_toio(unsigned long to, unsigned long from, unsigned long len)
+{
+	to += isa_slot_offset;
+	memcpy((void *)to, (void *)from, len);
+}
diff --git a/arch/mips/deskstation/setup.c b/arch/mips/deskstation/setup.c
index c2451e855..8b70a656c 100644
--- a/arch/mips/deskstation/setup.c
+++ b/arch/mips/deskstation/setup.c
@@ -5,10 +5,10 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 1996 by Ralf Baechle
+ * Copyright (C) 1996, 1997 by Ralf Baechle
  */
-#include <asm/ptrace.h>
 #include <linux/config.h>
+#include <linux/init.h>
 #include <linux/ioport.h>
 #include <linux/sched.h>
 #include <linux/interrupt.h>
@@ -16,6 +16,7 @@
 #include <asm/bootinfo.h>
 #include <asm/io.h>
 #include <asm/irq.h>
+#include <asm/ptrace.h>
 #include <asm/mipsregs.h>
 #include <asm/reboot.h>
 #include <asm/vector.h>
@@ -43,8 +44,7 @@ extern void deskstation_machine_power_off(void);
 unsigned long mips_dma_cache_size = 0;
 unsigned long mips_dma_cache_base = KSEG0;
 
-static void
-tyne_irq_setup(void)
+__initfunc(static void tyne_irq_setup(void))
 {
 	set_except_vector(0, deskstation_handle_int);
 	/* set the clock to 100 Hz */
@@ -58,8 +58,7 @@ tyne_irq_setup(void)
 #endif
 
 #ifdef CONFIG_DESKSTATION_RPC44
-static void
-rpc44_irq_setup(void)
+__initfunc(static void rpc44_irq_setup(void))
 {
 	/*
 	 * For the moment just steal the TYNE support.  In the
@@ -77,8 +76,7 @@ rpc44_irq_setup(void)
 }
 #endif
 
-void
-deskstation_setup(void)
+__initfunc(void deskstation_setup(void))
 {
 	switch(mips_machtype) {
 #ifdef CONFIG_DESKSTATION_TYNE
diff --git a/arch/mips/jazz/.cvsignore b/arch/mips/jazz/.cvsignore
new file mode 100644
index 000000000..4671378ae
--- /dev/null
+++ b/arch/mips/jazz/.cvsignore
@@ -0,0 +1 @@
+.depend
diff --git a/arch/mips/jazz/io.c b/arch/mips/jazz/io.c
new file mode 100644
index 000000000..a151b99fe
--- /dev/null
+++ b/arch/mips/jazz/io.c
@@ -0,0 +1,136 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Low level I/O functions for Jazz family machines.
+ *
+ * Copyright (C) 1997 by Ralf Baechle.
+ */
+#include <linux/string.h>
+#include <asm/mipsconfig.h>
+#include <asm/addrspace.h>
+#include <asm/system.h>
+#include <asm/spinlock.h>
+#include <asm/jazz.h>
+
+/*
+ * Map an 16mb segment of the EISA address space to 0xe3000000;
+ */
+static inline void map_eisa_address(unsigned long address)
+{
+  /* XXX */
+  /* We've got an wired entry in the TLB.  We just need to modify it.
+     fast and clean.  But since we want to get rid of wired entries
+     things are a little bit more complicated ... */
+}
+
+static unsigned char jazz_readb(unsigned long addr)
+{
+	unsigned char res;
+
+	map_eisa_address(addr);
+	addr &= 0xffffff;
+	res = *(volatile unsigned char *) (JAZZ_EISA_BASE + addr);
+
+	return res;
+}
+
+static unsigned short jazz_readw(unsigned long addr)
+{
+	unsigned short res;
+
+	map_eisa_address(addr);
+	addr &= 0xffffff;
+	res = *(volatile unsigned char *) (JAZZ_EISA_BASE + addr);
+
+	return res;
+}
+
+static unsigned int jazz_readl(unsigned long addr)
+{
+	unsigned int res;
+
+	map_eisa_address(addr);
+	addr &= 0xffffff;
+	res = *(volatile unsigned char *) (JAZZ_EISA_BASE + addr);
+
+	return res;
+}
+
+static void jazz_writeb(unsigned char val, unsigned long addr)
+{
+	map_eisa_address(addr);
+	addr &= 0xffffff;
+	*(volatile unsigned char *) (JAZZ_EISA_BASE + addr) = val;
+}
+
+static void jazz_writew(unsigned short val, unsigned long addr)
+{
+	map_eisa_address(addr);
+	addr &= 0xffffff;
+	*(volatile unsigned char *) (JAZZ_EISA_BASE + addr) = val;
+}
+
+static void jazz_writel(unsigned int val, unsigned long addr)
+{
+	map_eisa_address(addr);
+	addr &= 0xffffff;
+	*(volatile unsigned char *) (JAZZ_EISA_BASE + addr) = val;
+}
+
+static void jazz_memset_io(unsigned long addr, int val, unsigned long len)
+{
+	unsigned long waddr;
+
+	waddr = JAZZ_EISA_BASE | (addr & 0xffffff);
+	while(len) {
+		unsigned long fraglen;
+
+		fraglen = (~addr + 1) & 0xffffff;
+		fraglen = (fraglen < len) ? fraglen : len;
+		map_eisa_address(addr);
+		memset((char *)waddr, val, fraglen);
+		addr += fraglen;
+		waddr = waddr + fraglen - 0x1000000;
+		len -= fraglen;
+	}
+}
+
+static void jazz_memcpy_fromio(unsigned long to, unsigned long from, unsigned long len)
+{
+	unsigned long waddr;
+
+	waddr = JAZZ_EISA_BASE | (from & 0xffffff);
+	while(len) {
+		unsigned long fraglen;
+
+		fraglen = (~from + 1) & 0xffffff;
+		fraglen = (fraglen < len) ? fraglen : len;
+		map_eisa_address(from);
+		memcpy((void *)to, (void *)waddr, fraglen);
+		to += fraglen;
+		from += fraglen;
+		waddr = waddr + fraglen - 0x1000000;
+		len -= fraglen;
+	}
+}
+
+static void jazz_memcpy_toio(unsigned long to, unsigned long from, unsigned long len)
+{
+	unsigned long waddr;
+
+	waddr = JAZZ_EISA_BASE | (to & 0xffffff);
+	while(len) {
+		unsigned long fraglen;
+
+		fraglen = (~to + 1) & 0xffffff;
+		fraglen = (fraglen < len) ? fraglen : len;
+		map_eisa_address(to);
+		memcpy((char *)to + JAZZ_EISA_BASE, (void *)from, fraglen);
+		to += fraglen;
+		from += fraglen;
+		waddr = waddr + fraglen - 0x1000000;
+		len -= fraglen;
+	}
+}
diff --git a/arch/mips/jazz/setup.c b/arch/mips/jazz/setup.c
index 1b137cf85..805efa821 100644
--- a/arch/mips/jazz/setup.c
+++ b/arch/mips/jazz/setup.c
@@ -5,14 +5,15 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 1996 by Ralf Baechle
+ * Copyright (C) 1996, 1997 by Ralf Baechle
  */
-#include <asm/ptrace.h>
+#include <linux/init.h>
 #include <linux/ioport.h>
 #include <linux/sched.h>
 #include <linux/interrupt.h>
 #include <asm/irq.h>
 #include <asm/jazz.h>
+#include <asm/ptrace.h>
 #include <asm/reboot.h>
 #include <asm/vector.h>
 #include <asm/io.h>
@@ -35,8 +36,7 @@ extern void jazz_machine_restart(char *command);
 extern void jazz_machine_halt(void);
 extern void jazz_machine_power_off(void);
 
-static void
-jazz_irq_setup(void)
+__initfunc(static void jazz_irq_setup(void))
 {
         set_except_vector(0, jazz_handle_int);
 	r4030_write_reg16(JAZZ_IO_IRQ_ENABLE,
@@ -55,8 +55,7 @@ jazz_irq_setup(void)
 	setup_x86_irq(2, &irq2);
 }
 
-void
-jazz_setup(void)
+__initfunc(void jazz_setup(void))
 {
 	irq_setup = jazz_irq_setup;
 	fd_cacheflush = jazz_fd_cacheflush;
diff --git a/arch/mips/kernel/Makefile b/arch/mips/kernel/Makefile
index b251d6e1a..21dd3610b 100644
--- a/arch/mips/kernel/Makefile
+++ b/arch/mips/kernel/Makefile
@@ -9,7 +9,7 @@
 .S.o:
 	$(CC) $(CFLAGS) -c $< -o $*.o
 
-all:	kernel.o head.o
+all:	kernel.o head.o init_task.o
 EXTRA_ASFLAGS = -mips3 -mcpu=r4000
 O_TARGET := kernel.o
 O_OBJS	:= branch.o process.o signal.o entry.o traps.o ptrace.o vm86.o \
diff --git a/arch/mips/kernel/entry.S b/arch/mips/kernel/entry.S
index 8bb8fb41c..682e00cda 100644
--- a/arch/mips/kernel/entry.S
+++ b/arch/mips/kernel/entry.S
@@ -14,9 +14,11 @@
  * and faults that can result in a task-switch. The ISA dependent TLB
  * code is in arch/mips/<ISA-level>/<cputype>.S
  */
+#include <linux/config.h>
 #include <linux/sys.h>
 
 #include <asm/asm.h>
+#include <asm/current.h>
 #include <asm/errno.h>
 #include <asm/mipsregs.h>
 #include <asm/mipsconfig.h>
@@ -53,7 +55,6 @@ reschedule:
 		 nop
 
 EXPORT(ret_from_sys_call)
-
 		 lw	t0,bh_mask
 		lw	t1,bh_active		# unused delay slot
 		and	t0,t1
@@ -64,7 +65,7 @@ EXPORT(ret_from_sys_call)
 		beqz	t1,return		# -> yes
 		 lw	t1,need_resched
 		bnez	t1,reschedule
-		 lw	s0,current_set
+		 GET_CURRENT(s0)
 		lw	t0,task
 		lw	a0,TASK_BLOCKED(s0)
 
diff --git a/arch/mips/kernel/head.S b/arch/mips/kernel/head.S
index a957e16bd..4c3e5bd52 100644
--- a/arch/mips/kernel/head.S
+++ b/arch/mips/kernel/head.S
@@ -22,6 +22,15 @@
 #include <asm/bootinfo.h>
 #include <asm/cpu.h>
 
+/*
+ * Get current task pointer
+ */
+#define GET_CURRENT(reg)			\
+	lui	reg, %hi(kernelsp);		\
+	lw	reg, %lo(kernelsp)(reg);	\
+	ori	reg, 8191;			\
+	xori	reg, 8191
+
 	.text
 	/*
 	 * Reserved space for exception handlers.
@@ -43,8 +52,7 @@
 	LEAF(except_vec0_r4000)
 	.set	mips3
 	mfc0	k0, CP0_BADVADDR		# Get faulting address
-	lui	k1, %hi(current_set)
-	lw	k1, %lo(current_set)(k1)	# get current task ptr
+	GET_CURRENT(k1)				# get current task ptr
 	srl	k0, k0, 22			# get pgd only bits
 	lw	k1, THREAD_PGDIR(k1)		# get task pg_dir
 	sll	k0, k0, 2
@@ -71,8 +79,7 @@
 	LEAF(except_vec0_r4600)
 	.set	mips3
 	mfc0	k0, CP0_BADVADDR
-	lui	k1, %hi(current_set)
-	lw	k1, %lo(current_set)(k1)
+	GET_CURRENT(k1)				# get current task ptr
 	srl	k0, k0, 22
 	lw	k1, THREAD_PGDIR(k1)
 	sll	k0, k0, 2
@@ -98,8 +105,7 @@
 	LEAF(except_vec0_r45k_bvahwbug)
 	.set	mips3
 	mfc0	k0, CP0_BADVADDR
-	lui	k1, %hi(current_set)
-	lw	k1, %lo(current_set)(k1)
+	GET_CURRENT(k1)				# get current task ptr
 	srl	k0, k0, 22
 	lw	k1, THREAD_PGDIR(k1)
 	sll	k0, k0, 2
@@ -129,8 +135,7 @@
 	LEAF(except_vec0_r4k_mphwbug)
 	.set	mips3
 	mfc0	k0, CP0_BADVADDR
-	lui	k1, %hi(current_set)
-	lw	k1, %lo(current_set)(k1)
+	GET_CURRENT(k1)				# get current task ptr
 	srl	k0, k0, 22
 	lw	k1, THREAD_PGDIR(k1)
 	sll	k0, k0, 2
@@ -160,8 +165,7 @@
 	LEAF(except_vec0_r4k_250MHZhwbug)
 	.set	mips3
 	mfc0	k0, CP0_BADVADDR
-	lui	k1, %hi(current_set)
-	lw	k1, %lo(current_set)(k1)
+	GET_CURRENT(k1)				# get current task ptr
 	srl	k0, k0, 22
 	lw	k1, THREAD_PGDIR(k1)
 	sll	k0, k0, 2
@@ -191,8 +195,7 @@
 	LEAF(except_vec0_r4k_MP250MHZhwbug)
 	.set	mips3
 	mfc0	k0, CP0_BADVADDR
-	lui	k1, %hi(current_set)
-	lw	k1, %lo(current_set)(k1)
+	GET_CURRENT(k1)				# get current task ptr
 	srl	k0, k0, 22
 	lw	k1, THREAD_PGDIR(k1)
 	sll	k0, k0, 2
@@ -224,8 +227,7 @@
 	LEAF(except_vec0_r2300)
 	.set	mips1
 	mfc0	k0, CP0_BADVADDR
-	lui	k1, %hi(current_set)
-	lw	k1, %lo(current_set)(k1)
+	GET_CURRENT(k1)				# get current task ptr
 	srl	k0, k0, 22
 	lw	k1, THREAD_PGDIR(k1)
 	sll	k0, k0, 2
@@ -403,9 +405,11 @@ probe_done:
 
 	/*
 	 * Stack for kernel and init
+	 *
+	 * Kernelsp will never be referenced for process 0.
 	 */
-9:	la	sp, init_user_stack+(KERNEL_STACK_SIZE-4*SZREG)
-	la	t0, init_kernel_stack+(KERNEL_STACK_SIZE)
+9:	la	sp, init_task_union+(KERNEL_STACK_SIZE-4*SZREG)
+	la	t0, init_task_union+(KERNEL_STACK_SIZE)
 	sw	t0, kernelsp
 
 	/* Disable coprocessors */
@@ -745,9 +749,13 @@ map0_sni_rm200_pci:
 
 			.org	0x6000
 
+			/*
+			 * init_task_union follows here in the .text segment.
+			 * Keep this aligned to a 8kb boundary!
+			 */
+			.data
 			EXPORT(cache_error_buffer)
 			.fill	32*4,1,0
 
-			.data
 			EXPORT(kernelsp)
 			PTR	0
diff --git a/arch/mips/kernel/init_task.c b/arch/mips/kernel/init_task.c
new file mode 100644
index 000000000..cc0a19231
--- /dev/null
+++ b/arch/mips/kernel/init_task.c
@@ -0,0 +1,22 @@
+#include <linux/mm.h>
+#include <linux/sched.h>
+
+#include <asm/pgtable.h>
+
+static struct vm_area_struct init_mmap = INIT_MMAP;
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS;
+struct mm_struct init_mm = INIT_MM;
+
+/*
+ * Initial task structure.
+ *
+ * We need to make sure that this is 8192-byte aligned due to the
+ * way process stacks are handled. This is done by making sure
+ * the linker maps this in the .text segment right after head.S,
+ * and making head.S ensure the proper alignment.
+ *
+ * The things we do for performance..
+ */
+union task_union init_task_union __attribute__((__section__(".text"))) = { INIT_TASK };
diff --git a/arch/mips/kernel/irix5sys.h b/arch/mips/kernel/irix5sys.h
index a20e619e3..be57dc1f5 100644
--- a/arch/mips/kernel/irix5sys.h
+++ b/arch/mips/kernel/irix5sys.h
@@ -105,7 +105,7 @@ SYS(irix_sgikopt, 3)			/* 1083  sys_sgikopt()	      DC*/
 SYS(sys_sysfs, 3)			/* 1084  sysfs()	      ?V*/
 SYS(irix_unimp, 0)			/* 1085  XXX sys_getmsg()     DC*/
 SYS(irix_unimp, 0)			/* 1086  XXX sys_putmsg()     DC*/
-SYS(irix_poll, 3)			/* 1087  sys_poll()	       V*/
+SYS(sys_poll, 3)			/* 1087  poll()	               V*/
 SYS(irix_sigreturn, 0)			/* 1088  sigreturn()	      ?V*/
 SYS(sys_accept, 3)			/* 1089  accept()	       V*/
 SYS(sys_bind, 3)			/* 1090  bind()		       V*/
diff --git a/arch/mips/kernel/irixelf.c b/arch/mips/kernel/irixelf.c
index 46345b308..d994155d0 100644
--- a/arch/mips/kernel/irixelf.c
+++ b/arch/mips/kernel/irixelf.c
@@ -1,4 +1,4 @@
-/* $Id: irixelf.c,v 1.8 1996/08/24 03:52:25 dm Exp $
+/*
  * irixelf.c: Code to load IRIX ELF executables which conform to
  *            the MIPS ABI.
  *
@@ -17,6 +17,7 @@
 #include <linux/mman.h>
 #include <linux/a.out.h>
 #include <linux/errno.h>
+#include <linux/init.h>
 #include <linux/signal.h>
 #include <linux/binfmts.h>
 #include <linux/string.h>
@@ -720,16 +721,16 @@ static inline int do_load_irix_binary(struct linux_binprm * bprm,
 	sys_close(elf_exec_fileno);
 	current->personality = PER_IRIX32;
 
-	if (current->exec_domain && current->exec_domain->use_count)
-		(*current->exec_domain->use_count)--;
-	if (current->binfmt && current->binfmt->use_count)
-		(*current->binfmt->use_count)--;
+	if (current->exec_domain && current->exec_domain->module)
+		__MOD_DEC_USE_COUNT(current->exec_domain->module);
+	if (current->binfmt && current->binfmt->module)
+		__MOD_DEC_USE_COUNT(current->binfmt->module);
 	current->exec_domain = lookup_exec_domain(current->personality);
 	current->binfmt = &irix_format;
-	if (current->exec_domain && current->exec_domain->use_count)
-		(*current->exec_domain->use_count)++;
-	if (current->binfmt && current->binfmt->use_count)
-		(*current->binfmt->use_count)++;
+	if (current->exec_domain && current->exec_domain->module)
+		__MOD_INC_USE_COUNT(current->exec_domain->module);
+	if (current->binfmt && current->binfmt->module)
+		__MOD_INC_USE_COUNT(current->binfmt->module);
 
 	current->suid = current->euid = current->fsuid = bprm->e_uid;
 	current->sgid = current->egid = current->fsgid = bprm->e_gid;
@@ -1180,20 +1181,20 @@ static int irix_core_dump(long signr, struct pt_regs * regs)
 	notes[0].datasz = sizeof(prstatus);
 	notes[0].data = &prstatus;
 	prstatus.pr_info.si_signo = prstatus.pr_cursig = signr;
-	copy_sigbits32(&prstatus.pr_sigpend, current->signal);
-	copy_sigbits32(&prstatus.pr_sighold, current->blocked);
+	prstatus.pr_sigpend = current->signal;
+	prstatus.pr_sighold = current->blocked;
 	psinfo.pr_pid = prstatus.pr_pid = current->pid;
 	psinfo.pr_ppid = prstatus.pr_ppid = current->p_pptr->pid;
 	psinfo.pr_pgrp = prstatus.pr_pgrp = current->pgrp;
 	psinfo.pr_sid = prstatus.pr_sid = current->session;
-	prstatus.pr_utime.tv_sec = CT_TO_SECS(current->utime);
-	prstatus.pr_utime.tv_usec = CT_TO_USECS(current->utime);
-	prstatus.pr_stime.tv_sec = CT_TO_SECS(current->stime);
-	prstatus.pr_stime.tv_usec = CT_TO_USECS(current->stime);
-	prstatus.pr_cutime.tv_sec = CT_TO_SECS(current->cutime);
-	prstatus.pr_cutime.tv_usec = CT_TO_USECS(current->cutime);
-	prstatus.pr_cstime.tv_sec = CT_TO_SECS(current->cstime);
-	prstatus.pr_cstime.tv_usec = CT_TO_USECS(current->cstime);
+	prstatus.pr_utime.tv_sec = CT_TO_SECS(current->times.tms_utime);
+	prstatus.pr_utime.tv_usec = CT_TO_USECS(current->times.tms_utime);
+	prstatus.pr_stime.tv_sec = CT_TO_SECS(current->times.tms_stime);
+	prstatus.pr_stime.tv_usec = CT_TO_USECS(current->times.tms_stime);
+	prstatus.pr_cutime.tv_sec = CT_TO_SECS(current->times.tms_cutime);
+	prstatus.pr_cutime.tv_usec = CT_TO_USECS(current->times.tms_cutime);
+	prstatus.pr_cstime.tv_sec = CT_TO_SECS(current->times.tms_cstime);
+	prstatus.pr_cstime.tv_usec = CT_TO_USECS(current->times.tms_cstime);
 	if (sizeof(elf_gregset_t) != sizeof(struct pt_regs)) {
 		printk("sizeof(elf_gregset_t) (%d) != sizeof(struct pt_regs) "
 		       "(%d)\n", sizeof(elf_gregset_t), sizeof(struct pt_regs));
@@ -1339,13 +1340,15 @@ static int irix_core_dump(long signr, struct pt_regs * regs)
 	return has_dumped;
 }
 
-int init_irix_binfmt(void) {
+__initfunc(int init_irix_binfmt(void))
+{
 	return register_binfmt(&irix_format);
 }
 
 #ifdef MODULE
 
-int init_module(void) {
+int init_module(void)
+{
 	/* Install the COFF, ELF and XOUT loaders.
 	 * N.B. We *rely* on the table being the right size with the
 	 * right number of free slots...
@@ -1354,7 +1357,8 @@ int init_module(void) {
 }
 
 
-void cleanup_module( void) {
+void cleanup_module( void)
+{
 	/* Remove the IRIX ELF loaders. */
 	unregister_binfmt(&irix_format);
 }
diff --git a/arch/mips/kernel/irixioctl.c b/arch/mips/kernel/irixioctl.c
index 3405904fa..e0230f9f0 100644
--- a/arch/mips/kernel/irixioctl.c
+++ b/arch/mips/kernel/irixioctl.c
@@ -8,6 +8,8 @@
 #include <linux/sched.h>
 #include <linux/fs.h>
 #include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
 #include <linux/tty.h>
 
 #include <asm/uaccess.h>
diff --git a/arch/mips/kernel/irixsig.c b/arch/mips/kernel/irixsig.c
index 6a5636c76..b2d7cbe49 100644
--- a/arch/mips/kernel/irixsig.c
+++ b/arch/mips/kernel/irixsig.c
@@ -8,6 +8,8 @@
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/errno.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
 #include <linux/time.h>
 
 #include <asm/ptrace.h>
@@ -162,7 +164,10 @@ asmlinkage int do_irix_signal(unsigned long oldmask, struct pt_regs * regs)
 			case SIGCONT: case SIGCHLD: case SIGWINCH:
 				continue;
 
-			case SIGSTOP: case SIGTSTP: case SIGTTIN: case SIGTTOU:
+			case SIGTSTP: case SIGTTIN: case SIGTTOU:
+				if (is_orphaned_pgrp(current->pgrp))
+					continue;
+			case SIGSTOP:
 				if (current->flags & PF_PTRACED)
 					continue;
 				current->state = TASK_STOPPED;
@@ -175,15 +180,19 @@ asmlinkage int do_irix_signal(unsigned long oldmask, struct pt_regs * regs)
 
 			case SIGQUIT: case SIGILL: case SIGTRAP:
 			case SIGIOT: case SIGFPE: case SIGSEGV: case SIGBUS:
+				lock_kernel();
 				if (current->binfmt && current->binfmt->core_dump) {
 					if (current->binfmt->core_dump(signr, regs))
 						signr |= 0x80;
 				}
+				unlock_kernel();
 				/* fall through */
 			default:
 				current->signal |= _S(signr & 0x7f);
 				current->flags |= PF_SIGNALED;
+				lock_kernel(); /* 8-( */
 				do_exit(signr);
+				unlock_kernel();
 			}
 		}
 		/*
@@ -314,7 +323,7 @@ static inline void check_pending(int signum)
 	spin_lock(&current->sigmask_lock);
 	if (p->sa_handler == SIG_IGN) {
 		current->signal &= ~_S(signum);
-	} else if if (p->sa_handler == SIG_DFL) {
+	} else if (p->sa_handler == SIG_DFL) {
 		if (signum != SIGCONT && signum != SIGCHLD && signum != SIGWINCH)
 			return;
 		current->signal &= ~_S(signum);
@@ -347,11 +356,9 @@ asmlinkage int irix_sigaction(int sig, struct sigact_irix5 *new,
 		if(sig == SIGKILL || sig == SIGSTOP) {
 			return -EINVAL;
 		}
-		new_sa.sa_flags = new->flags;
-		new_sa.sa_handler = (__sighandler_t) new->handler;
-		new_sa.sa_mask.__sigbits[1] = new_sa.sa_mask.__sigbits[2] =
-			new_sa.sa_mask.__sigbits[3] = 0;
-		new_sa.sa_mask.__sigbits[0] = new->sigset[0];
+		__get_user(new_sa.sa_flags, &new->flags);
+		__get_user(new_sa.sa_handler, &(__sighandler_t) new->handler);
+		__get_user(new_sa.sa_mask, &new->sigset[0]);
 
 		if(new_sa.sa_handler != SIG_DFL && new_sa.sa_handler != SIG_IGN) {
 			res = verify_area(VERIFY_READ, new_sa.sa_handler, 1);
@@ -368,19 +375,22 @@ asmlinkage int irix_sigaction(int sig, struct sigact_irix5 *new,
 		int res = verify_area(VERIFY_WRITE, old, sizeof(*old));
 		if(res)
 			return res;
-		old->flags = p->sa_flags;
-		old->handler = (void *) p->sa_handler;
-		old->sigset[1] = old->sigset[2] = old->sigset[3] = 0;
-		old->sigset[0] = p->sa_mask.__sigbits[0];
-		old->_unused0[0] = old->_unused0[1] = 0;
+		__put_user(p->sa_flags, &old->flags);
+		__put_user(p->sa_handler, &old->handler);
+		__put_user(p->sa_mask, &old->sigset[0]);
+		__put_user(0, &old->sigset[1]);
+		__put_user(0, &old->sigset[2]);
+		__put_user(0, &old->sigset[3]);
+		__put_user(0, &old->_unused0[0]);
+		__put_user(0, &old->_unused0[1]);
 	}
-
 	if(new) {
 		spin_lock_irq(&current->sig->siglock);
 		*p = new_sa;
 		check_pending(sig);
 		spin_unlock_irq(&current->sig->siglock);
 	}
+
 	return 0;
 }
 
@@ -640,14 +650,14 @@ repeat:
 				__put_user(p->pid, &info->stuff.procinfo.pid);
 				__put_user((p->exit_code >> 8) & 0xff,
 				           &info->stuff.procinfo.procdata.child.status);
-				__put_user(p->utime, &info->stuff.procinfo.procdata.child.utime);
-				__put_user(p->stime, &info->stuff.procinfo.procdata.child.stime);
+				__put_user(p->times.tms_utime, &info->stuff.procinfo.procdata.child.utime);
+				__put_user(p->times.tms_stime, &info->stuff.procinfo.procdata.child.stime);
 				p->exit_code = 0;
 				retval = 0;
 				goto end_waitsys;
 			case TASK_ZOMBIE:
-				current->cutime += p->utime + p->cutime;
-				current->cstime += p->stime + p->cstime;
+				current->times.tms_cutime += p->times.tms_utime + p->times.tms_cutime;
+				current->times.tms_cstime += p->times.tms_stime + p->times.tms_cstime;
 				if (ru != NULL)
 					getrusage(p, RUSAGE_BOTH, ru);
 				__put_user(SIGCHLD, &info->sig);
@@ -655,9 +665,9 @@ repeat:
 				__put_user(p->pid, &info->stuff.procinfo.pid);
 				__put_user((p->exit_code >> 8) & 0xff,
 				           &info->stuff.procinfo.procdata.child.status);
-				__put_user(p->utime,
+				__put_user(p->times.tms_utime,
 				           &info->stuff.procinfo.procdata.child.utime);
-				__put_user(p->stime,
+				__put_user(p->times.tms_stime,
 				           &info->stuff.procinfo.procdata.child.stime);
 				retval = 0;
 				if (p->p_opptr != p->p_pptr) {
@@ -855,6 +865,8 @@ asmlinkage int irix_sigaltstack(struct irix_sigaltstack *new,
 out:
 	error = 0;
 	unlock_kernel();
+
+	return error;
 }
 
 struct irix_procset {
diff --git a/arch/mips/kernel/irq.c b/arch/mips/kernel/irq.c
index a78bc3417..0896ed1c7 100644
--- a/arch/mips/kernel/irq.c
+++ b/arch/mips/kernel/irq.c
@@ -13,6 +13,7 @@
  */
 #include <linux/config.h>
 #include <linux/errno.h>
+#include <linux/init.h>
 #include <linux/kernel_stat.h>
 #include <linux/signal.h>
 #include <linux/sched.h>
@@ -154,11 +155,6 @@ asmlinkage void do_IRQ(int irq, struct pt_regs * regs)
 	irq_enter(cpu, irq);
 	kstat.interrupts[irq]++;
 
-#ifdef CONFIG_SGI
-	prom_printf("Got irq %d, press a key.", irq);
-	prom_getchar();
-	romvec->imode();
-#endif
 	/* slow interrupts run with interrupts enabled */
 	sti();
 	action = *(irq + irq_action);
@@ -344,7 +340,7 @@ int probe_irq_off (unsigned long irqs)
 	return i;
 }
 
-void init_IRQ(void)
+__initfunc(void init_IRQ(void))
 {
 	int i;
 
diff --git a/arch/mips/kernel/pci.c b/arch/mips/kernel/pci.c
index e521ecdd9..5e71233af 100644
--- a/arch/mips/kernel/pci.c
+++ b/arch/mips/kernel/pci.c
@@ -7,6 +7,7 @@
  */
 #include <linux/bios32.h>
 #include <linux/config.h>
+#include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/pci.h>
 #include <linux/types.h>
@@ -17,7 +18,8 @@
 /*
  * BIOS32 replacement.
  */
-unsigned long pcibios_init(unsigned long memory_start, unsigned long memory_end)
+__initfunc(unsigned long pcibios_init(unsigned long memory_start,
+                                      unsigned long memory_end))
 {
 	return memory_start;
 }
@@ -112,7 +114,8 @@ const char *pcibios_strerror (int error)
  * specific implementation.
  */
 unsigned long (*_pcibios_init)(unsigned long memory_start, unsigned long memory_end);
-unsigned long pcibios_init(unsigned long memory_start, unsigned long memory_end)
+__initfunc(unsigned long pcibios_init(unsigned long memory_start,
+                                      unsigned long memory_end))
 {
 	return _pcibios_init ? _pcibios_init(memory_start, memory_end)
 	                     : memory_start;
diff --git a/arch/mips/kernel/process.c b/arch/mips/kernel/process.c
index f8b10bdea..08dd13c6c 100644
--- a/arch/mips/kernel/process.c
+++ b/arch/mips/kernel/process.c
@@ -64,16 +64,23 @@ void release_thread(struct task_struct *dead_task)
 {
 }
 
+#define roundup(val, rnd) ({ \
+	unsigned _v = val; \
+	unsigned long _r = rnd; \
+	_v = (_v + _r - 1) & ~(_r - 1); \
+	_v; \
+})
+
 int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
                  struct task_struct * p, struct pt_regs * regs)
 {
 	struct pt_regs * childregs;
 	long childksp;
 
-	childksp = p->kernel_stack_page + KERNEL_STACK_SIZE - 8;
+	childksp = roundup((unsigned long)p, KERNEL_STACK_SIZE) - 8;
 
 	/* set up new TSS. */
-	childregs = ((struct pt_regs *) (p->kernel_stack_page + PAGE_SIZE)) - 1;
+	childregs = ((struct pt_regs *) ((unsigned long)p + KERNEL_STACK_SIZE)) - 1;
 	*childregs = *regs;
 	childregs->regs[7] = 0;	/* Clear error flag */
 	if(current->personality == PER_LINUX) {
diff --git a/arch/mips/kernel/ptrace.c b/arch/mips/kernel/ptrace.c
index 41deb8f18..e61911549 100644
--- a/arch/mips/kernel/ptrace.c
+++ b/arch/mips/kernel/ptrace.c
@@ -18,18 +18,6 @@
 #include <asm/page.h>
 #include <asm/system.h>
 
-/* change a pid into a task struct. */
-static inline struct task_struct * get_task(int pid)
-{
-	int i;
-
-	for (i = 1; i < NR_TASKS; i++) {
-		if (task[i] != NULL && (task[i]->pid == pid))
-			return task[i];
-	}
-	return NULL;
-}
-
 /*
  * This routine gets a long from any process space by following the page
  * tables. NOTE! You should check that the long isn't on a page boundary,
@@ -47,7 +35,7 @@ static unsigned long get_long(struct task_struct * tsk,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (pgd_none(*pgdir)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -57,7 +45,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -67,7 +55,7 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
@@ -101,7 +89,7 @@ static void put_long(struct task_struct *tsk,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (!pgd_present(*pgdir)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -111,7 +99,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -121,12 +109,12 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
 	if (!pte_write(*pgtable)) {
-		do_wp_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 /* this is a hack for non-kernel-mapped video buffers and similar */
@@ -280,7 +268,7 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 		res = -EPERM;
 		goto out;
 	}
-	if (!(child = get_task(pid))) {
+	if (!(child = find_task_by_pid(pid))) {
 		res = -ESRCH;
 		goto out;
 	}
diff --git a/arch/mips/kernel/r2300_switch.S b/arch/mips/kernel/r2300_switch.S
index 83190514b..901871a31 100644
--- a/arch/mips/kernel/r2300_switch.S
+++ b/arch/mips/kernel/r2300_switch.S
@@ -9,6 +9,7 @@
 #include <asm/asm.h>
 #include <asm/bootinfo.h>
 #include <asm/cachectl.h>
+#include <asm/current.h>
 #include <asm/fpregdef.h>
 #include <asm/mipsconfig.h>
 #include <asm/mipsregs.h>
@@ -31,8 +32,7 @@ MODE_ALIAS	=	0x00e0			# uncachable, dirty, valid
  */
 	.align	5
 	LEAF(r2300_resume)
-	lui	t5, %hi(current_set)
-	lw	t0, %lo(current_set)(t5)
+	GET_CURRENT(t0)
 	mfc0	t1,CP0_STATUS		# Save status register
 	addu	t0,a1			# Add tss offset
 	sw	t1,THREAD_STATUS(t0)
@@ -50,7 +50,6 @@ MODE_ALIAS	=	0x00e0			# uncachable, dirty, valid
 1:
 	FPU_SAVE_16EVEN(t0, t1)
 2:
-	sw	a0,%lo(current_set)(t5)	# Switch current task
 	addu	a0,a1			# Add tss offset
 	lw	t0,THREAD_PGDIR(a0)	# Switch the root pointer
 	li	t1,TLB_ROOT		# get PFN
diff --git a/arch/mips/kernel/r4k_misc.S b/arch/mips/kernel/r4k_misc.S
index 432c65215..5c9ad4d84 100644
--- a/arch/mips/kernel/r4k_misc.S
+++ b/arch/mips/kernel/r4k_misc.S
@@ -10,6 +10,7 @@
 #include <asm/offset.h>
 #include <asm/bootinfo.h>
 #include <asm/cachectl.h>
+#include <asm/current.h>
 #include <asm/fpregdef.h>
 #include <asm/mipsconfig.h>
 #include <asm/mipsregs.h>
@@ -29,9 +30,8 @@
 	 */
 #define LOAD_PTE(pte, ptr) \
 	mfc0	pte, CP0_BADVADDR; \
-	lui	ptr, %hi(current_set); \
 	srl	pte, pte, 22; \
-	lw	ptr, %lo(current_set)(ptr); \
+	GET_CURRENT(ptr); \
 	sll	pte, pte, 2; \
 	lw	ptr, THREAD_PGDIR(ptr); \
 	addu	ptr, pte, ptr; \
diff --git a/arch/mips/kernel/r4k_switch.S b/arch/mips/kernel/r4k_switch.S
index 78ced5659..97e253028 100644
--- a/arch/mips/kernel/r4k_switch.S
+++ b/arch/mips/kernel/r4k_switch.S
@@ -1,4 +1,4 @@
-/* $Id: r4k_switch.S,v 1.8 1996/07/10 01:24:20 dm Exp $
+/*
  * r4k_switch.S: R4xx0 specific task switching code.
  *
  * Copyright (C) 1994, 1995, 1996 by Ralf Baechle and Andreas Busse
@@ -9,6 +9,7 @@
 #include <asm/asm.h>
 #include <asm/bootinfo.h>
 #include <asm/cachectl.h>
+#include <asm/current.h>
 #include <asm/fpregdef.h>
 #include <asm/mipsconfig.h>
 #include <asm/mipsregs.h>
@@ -24,8 +25,7 @@
 	.set	mips3
 	.align	5
 	LEAF(r4xx0_resume)
-	lui	t5, %hi(current_set)
-	lw	t0, %lo(current_set)(t5)
+	GET_CURRENT(t0)
 	mfc0	t1, CP0_STATUS
 	nop
 	sw	t1, THREAD_STATUS(t0)
@@ -43,7 +43,6 @@
 1:
 	FPU_SAVE_16EVEN(t0, t1)		# clobbers t1
 2:
-	sw	a0, %lo(current_set)(t5)
 	lw	a3, TASK_MM(a0)
 	lw	a2, THREAD_STATUS(a0)
 	lw	a3, MM_CONTEXT(a3)
diff --git a/arch/mips/kernel/setup.c b/arch/mips/kernel/setup.c
index 2e2b074f9..7616fa7c6 100644
--- a/arch/mips/kernel/setup.c
+++ b/arch/mips/kernel/setup.c
@@ -7,6 +7,7 @@
  */
 #include <linux/config.h>
 #include <linux/errno.h>
+#include <linux/init.h>
 #include <linux/ioport.h>
 #include <linux/sched.h>
 #include <linux/kernel.h>
@@ -123,17 +124,17 @@ void (*irq_setup)(void);
  */
 unsigned long isa_slot_offset;
 
-static void default_irq_setup(void)
+__initfunc(static void default_irq_setup(void))
 {
 	panic("Unknown machtype in init_IRQ");
 }
 
-static void default_fd_cacheflush(const void *addr, size_t size)
+__initfunc(static void default_fd_cacheflush(const void *addr, size_t size))
 {
 }
 
-void setup_arch(char **cmdline_p,
-                unsigned long * memory_start_p, unsigned long * memory_end_p)
+__initfunc(void setup_arch(char **cmdline_p,
+           unsigned long * memory_start_p, unsigned long * memory_end_p))
 {
 	unsigned long memory_end;
 	tag* atag;
diff --git a/arch/mips/kernel/signal.c b/arch/mips/kernel/signal.c
index 848f0742b..304dc6418 100644
--- a/arch/mips/kernel/signal.c
+++ b/arch/mips/kernel/signal.c
@@ -307,7 +307,10 @@ asmlinkage int do_signal(unsigned long oldmask, struct pt_regs * regs)
 			case SIGCONT: case SIGCHLD: case SIGWINCH:
 				continue;
 
-			case SIGSTOP: case SIGTSTP: case SIGTTIN: case SIGTTOU:
+			case SIGTSTP: case SIGTTIN: case SIGTTOU:
+				if (is_orphaned_pgrp(current->pgrp))
+					continue;
+			case SIGSTOP:
 				if (current->flags & PF_PTRACED)
 					continue;
 				current->state = TASK_STOPPED;
@@ -318,11 +321,15 @@ asmlinkage int do_signal(unsigned long oldmask, struct pt_regs * regs)
 				schedule();
 				continue;
 
-			case SIGIOT: case SIGFPE: case SIGSEGV: case SIGBUS:
+			case SIGQUIT: case SIGILL: case SIGTRAP:
+			case SIGABRT: case SIGFPE: case SIGSEGV:
+			case SIGBUS:
+				lock_kernel();
 				if (current->binfmt && current->binfmt->core_dump) {
 					if (current->binfmt->core_dump(signr, regs))
 						signr |= 0x80;
 				}
+				unlock_kernel();
 				/* fall through */
 			default:
 				spin_lock_irq(&current->sigmask_lock);
diff --git a/arch/mips/kernel/syscalls.h b/arch/mips/kernel/syscalls.h
index 4cf778acb..6c7d3988c 100644
--- a/arch/mips/kernel/syscalls.h
+++ b/arch/mips/kernel/syscalls.h
@@ -174,6 +174,7 @@ SYS(sys_mlock, 2)
 SYS(sys_munlock, 2)				/* 4155 */
 SYS(sys_mlockall, 1)
 SYS(sys_munlockall, 0)
+SYS(sys_nfsservctl, 3)
 SYS(sys_sched_setparam,2)
 SYS(sys_sched_getparam,2)
 SYS(sys_sched_setscheduler,3)			/* 4160 */
diff --git a/arch/mips/kernel/sysirix.c b/arch/mips/kernel/sysirix.c
index 9a4ddca3f..7917664fd 100644
--- a/arch/mips/kernel/sysirix.c
+++ b/arch/mips/kernel/sysirix.c
@@ -17,6 +17,8 @@
 #include <linux/elf.h>
 #include <linux/msg.h>
 #include <linux/shm.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
 #include <linux/utsname.h>
 
 #include <asm/ptrace.h>
@@ -24,26 +26,7 @@
 #include <asm/pgtable.h>
 #include <asm/uaccess.h>
 
-/* 2,000 lines of complete and utter shit coming up... */
-
-/* Utility routines. */
-static inline struct task_struct *find_process_by_pid(pid_t pid)
-{
-	struct task_struct *p, *q;
-
-	if (pid == 0)
-		p = current;
-	else {
-		p = 0;
-		for_each_task(q) {
-			if (q && q->pid == pid) {
-				p = q;
-				break;
-			}
-		}
-	}
-	return p;
-}
+/* 2,300 lines of complete and utter shit coming up... */
 
 /* The sysmp commands supported thus far. */
 #define MP_PGSIZE           14 /* Return system page size in v1. */
@@ -70,7 +53,6 @@ asmlinkage int irix_sysmp(struct pt_regs *regs)
 		break;
 	}
 
-out:
 	unlock_kernel();
 	return error;
 }
@@ -114,7 +96,7 @@ asmlinkage int irix_prctl(struct pt_regs *regs)
 
 		printk("irix_prctl[%s:%d]: Wants PR_ISBLOCKED\n",
 		       current->comm, current->pid);
-		task = find_process_by_pid(regs->regs[base + 5]);
+		task = find_task_by_pid(regs->regs[base + 5]);
 		if(!task) {
 			error = -ESRCH;
 			break;
@@ -233,7 +215,6 @@ asmlinkage int irix_prctl(struct pt_regs *regs)
 		break;
 	}
 
-out:
 	unlock_kernel();
 	return error;
 }
@@ -658,7 +639,6 @@ asmlinkage int irix_mount(char *dev_name, char *dir_name, unsigned long flags,
 	       dev_name, dir_name, flags, type, data, datalen);
 	ret = sys_mount(dev_name, dir_name, type, flags, data);
 
-out:
 	unlock_kernel();
 	return ret;
 }
@@ -781,22 +761,23 @@ asmlinkage int irix_setpgrp(int flags)
 	printk("returning %d\n", current->pgrp);
 #endif
 
-out:
 	unlock_kernel();
 	return error;
 }
 
 asmlinkage int irix_times(struct tms * tbuf)
 {
+	int error;
+
 	lock_kernel();
 	if (tbuf) {
-		int error = verify_area(VERIFY_WRITE,tbuf,sizeof *tbuf);
+		error = verify_area(VERIFY_WRITE,tbuf,sizeof *tbuf);
 		if (error)
 			goto out;
-		__put_user(current->utime,&tbuf->tms_utime);
-		__put_user(current->stime,&tbuf->tms_stime);
-		__put_user(current->cutime,&tbuf->tms_cutime);
-		__put_user(current->cstime,&tbuf->tms_cstime);
+		__put_user(current->times.tms_utime,&tbuf->tms_utime);
+		__put_user(current->times.tms_stime,&tbuf->tms_stime);
+		__put_user(current->times.tms_cutime,&tbuf->tms_cutime);
+		__put_user(current->times.tms_cstime,&tbuf->tms_cstime);
 	}
 	error = 0;
 
@@ -845,169 +826,6 @@ out:
 	return error;
 }
 
-/* sys_poll() support... */
-#define POLL_ROUND_UP(x,y) (((x)+(y)-1)/(y))
-
-#define POLLIN 1
-#define POLLPRI 2
-#define POLLOUT 4
-#define POLLERR 8
-#define POLLHUP 16
-#define POLLNVAL 32
-#define POLLRDNORM 64
-#define POLLWRNORM POLLOUT
-#define POLLRDBAND 128
-#define POLLWRBAND 256
-
-#define LINUX_POLLIN (POLLRDNORM | POLLRDBAND | POLLIN)
-#define LINUX_POLLOUT (POLLWRBAND | POLLWRNORM | POLLOUT)
-#define LINUX_POLLERR (POLLERR)
-
-static inline void free_wait(select_table * p)
-{
-	struct select_table_entry * entry = p->entry + p->nr;
-
-	while (p->nr > 0) {
-		p->nr--;
-		entry--;
-		remove_wait_queue(entry->wait_address,&entry->wait);
-	}
-}
-
-
-/* Copied directly from fs/select.c */
-static int check(int flag, select_table * wait, struct file * file)
-{
-	struct inode * inode;
-	struct file_operations *fops;
-	int (*select) (struct inode *, struct file *, int, select_table *);
-
-	inode = file->f_inode;
-	if ((fops = file->f_op) && (select = fops->select))
-		return select(inode, file, flag, wait)
-		    || (wait && select(inode, file, flag, NULL));
-	if (S_ISREG(inode->i_mode))
-		return 1;
-	return 0;
-}
-
-struct poll {
-	int fd;
-	short events;
-	short revents;
-};
-
-int irix_poll(struct poll * ufds, size_t nfds, int timeout)
-{
-        int i,j, count, fdcount, error, retflag;
-	struct poll * fdpnt;
-	struct poll * fds, *fds1;
-	select_table wait_table, *wait;
-	struct select_table_entry *entry;
-
-	lock_kernel();
-	if ((error = verify_area(VERIFY_READ, ufds, nfds*sizeof(struct poll))))
-		goto out;
-
-	if (nfds > NR_OPEN) {
-		error = -EINVAL;
-		goto out;
-	}
-
-	if (!(entry = (struct select_table_entry*)__get_free_page(GFP_KERNEL))
-	|| !(fds = (struct poll *)kmalloc(nfds*sizeof(struct poll), GFP_KERNEL))) {
-		error = -ENOMEM;
-		goto out;
-	}
-
-	copy_from_user(fds, ufds, nfds*sizeof(struct poll));
-
-	if (timeout < 0)
-		current->timeout = 0x7fffffff;
-	else {
-		current->timeout = jiffies + POLL_ROUND_UP(timeout, (1000/HZ));
-		if (current->timeout <= jiffies)
-			current->timeout = 0;
-	}
-
-	count = 0;
-	wait_table.nr = 0;
-	wait_table.entry = entry;
-	wait = &wait_table;
-
-	for(fdpnt = fds, j = 0; j < (int)nfds; j++, fdpnt++) {
-		i = fdpnt->fd;
-		fdpnt->revents = 0;
-		if (!current->files->fd[i] || !current->files->fd[i]->f_inode)
-			fdpnt->revents = POLLNVAL;
-	}
-repeat:
-	current->state = TASK_INTERRUPTIBLE;
-	for(fdpnt = fds, j = 0; j < (int)nfds; j++, fdpnt++) {
-		i = fdpnt->fd;
-
-		if(i < 0) continue;
-		if (!current->files->fd[i] || !current->files->fd[i]->f_inode) continue;
-
-		if ((fdpnt->events & LINUX_POLLIN)
-		&& check(SEL_IN, wait, current->files->fd[i])) {
-			retflag = 0;
-			if (fdpnt->events & POLLIN)
-				retflag = POLLIN;
-			if (fdpnt->events & POLLRDNORM)
-				retflag = POLLRDNORM;
-			fdpnt->revents |= retflag; 
-			count++;
-			wait = NULL;
-		}
-
-		if ((fdpnt->events & LINUX_POLLOUT) &&
-		check(SEL_OUT, wait, current->files->fd[i])) {
-			fdpnt->revents |= (LINUX_POLLOUT & fdpnt->events);
-			count++;
-			wait = NULL;
-		}
-
-		if (check(SEL_EX, wait, current->files->fd[i])) {
-			fdpnt->revents |= POLLHUP;
-			count++;
-			wait = NULL;
-		}
-	}
-
-	if ((current->signal & (~current->blocked))) {
-		error = -EINTR;
-		goto out;
-	}
-
-	wait = NULL;
-	if (!count && current->timeout > jiffies) {
-		schedule();
-		goto repeat;
-	}
-
-	free_wait(&wait_table);
-	free_page((unsigned long) entry);
-
-	/* OK, now copy the revents fields back to user space. */
-	fds1 = fds;
-	fdcount = 0;
-	for(i=0; i < (int)nfds; i++, ufds++, fds++) {
-		if (fds->revents) {
-			fdcount++;
-		}
-		put_user(fds->revents, &ufds->revents);
-	}
-	kfree(fds1);
-	current->timeout = 0;
-	current->state = TASK_RUNNING;
-	error = fdcount;
-
-out:
-	unlock_kernel();
-	return error;
-}
-
 asmlinkage unsigned long irix_gethostid(void)
 {
 	lock_kernel();
@@ -1263,7 +1081,6 @@ asmlinkage int irix_BSDsetpgrp(int pid, int pgrp)
 	printk("error = %d\n", error);
 #endif
 
-out:
 	unlock_kernel();
 	return error;
 }
@@ -1443,11 +1260,11 @@ asmlinkage int irix_lxstat(int version, char *filename, struct stat *statbuf)
 		if(error)
 			goto out;
 		error = irix_xstat32_xlate(&kb, statbuf);
-		goto error;
+		goto out;
 	}
 
 	case 3: {
-		sys_newlstat(filename, statbuf);
+		error = sys_newlstat(filename, statbuf);
 #ifdef DEBUG_XSTAT
 		printk("error[%d]\n", error);
 #endif
@@ -1456,7 +1273,7 @@ asmlinkage int irix_lxstat(int version, char *filename, struct stat *statbuf)
 
 		irix_xstat64_xlate(statbuf);
 		error = 0;
-		goto error;
+		goto out;
 	}
 
 	default:
@@ -1466,7 +1283,7 @@ asmlinkage int irix_lxstat(int version, char *filename, struct stat *statbuf)
 
 out:
 	unlock_kernel();
-	return errno;
+	return error;
 }
 
 extern asmlinkage int sys_newfstat(unsigned int fd, struct stat * statbuf);
@@ -2018,13 +1835,13 @@ out:
 
 asmlinkage int irix_getmountid(char *fname, unsigned long *midbuf)
 {
-	int errno;
+	int error;
 
 	lock_kernel();
 	printk("[%s:%d] irix_getmountid(%s, %p)\n",
 	       current->comm, current->pid, fname, midbuf);
-	errno = verify_area(VERIFY_WRITE, midbuf, (sizeof(unsigned long) * 4));
-	if(errno)
+	error = verify_area(VERIFY_WRITE, midbuf, (sizeof(unsigned long) * 4));
+	if(error)
 		goto out;
 
 	/*
@@ -2142,7 +1959,7 @@ asmlinkage int irix_ngetdents(unsigned int fd, void * dirent, unsigned int count
 	buf.error = 0;
 	error = file->f_op->readdir(file->f_inode, file, &buf, irix_filldir32);
 	if (error < 0)
-		goto out
+		goto out;
 	lastdirent = buf.previous;
 	if (!lastdirent) {
 		error = buf.error;
@@ -2431,6 +2248,8 @@ asmlinkage int irix_fcntl(int fd, int cmd, int arg)
 
 asmlinkage int irix_ulimit(int cmd, int arg)
 {
+	int retval;
+
 	lock_kernel();
 	switch(cmd) {
 	case 1:
diff --git a/arch/mips/kernel/time.c b/arch/mips/kernel/time.c
index aa4547456..fd57f6d0e 100644
--- a/arch/mips/kernel/time.c
+++ b/arch/mips/kernel/time.c
@@ -7,6 +7,7 @@
  * found in some MIPS systems.
  */
 #include <linux/errno.h>
+#include <linux/init.h>
 #include <linux/sched.h>
 #include <linux/kernel.h>
 #include <linux/param.h>
@@ -299,7 +300,7 @@ static struct irqaction irq0  = { timer_interrupt, 0, 0, "timer", NULL, NULL};
 
 void (*board_time_init)(struct irqaction *irq);
 
-void time_init(void)
+__initfunc(void time_init(void))
 {
 	unsigned int year, mon, day, hour, min, sec;
 	int i;
diff --git a/arch/mips/kernel/traps.c b/arch/mips/kernel/traps.c
index 17b1f9a28..cdbcacb43 100644
--- a/arch/mips/kernel/traps.c
+++ b/arch/mips/kernel/traps.c
@@ -17,6 +17,7 @@
  * Modified for R3000 by Paul M. Antoine, 1995, 1996
  */
 #include <linux/config.h>
+#include <linux/init.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
 #include <linux/smp_lock.h>
@@ -110,10 +111,8 @@ void show_registers(char * str, struct pt_regs * regs, long err)
 	/*
 	 * Dump the stack
 	 */
-	if (STACK_MAGIC != *(u32 *)current->kernel_stack_page)
-		printk("Corrupted stack page\n");
 	printk("Process %s (pid: %d, stackpage=%08lx)\nStack: ",
-		current->comm, current->pid, current->kernel_stack_page);
+		current->comm, current->pid, (unsigned long)current);
 	for(i=0;i<5;i++)
 		printk("%08x ", *sp++);
 	stack = (int *) sp;
@@ -389,7 +388,7 @@ void do_reserved(struct pt_regs *regs)
 	unlock_kernel();
 }
 
-static void watch_init(unsigned long cputype)
+static inline void watch_init(unsigned long cputype)
 {
 	switch(cputype) {
 	case CPU_R10000:
@@ -427,7 +426,7 @@ extern asmlinkage void r6000_restore_fp_context(struct sigcontext *sc);
 extern asmlinkage void r4xx0_resume(void *tsk);
 extern asmlinkage void r2300_resume(void *tsk);
 
-void trap_init(void)
+__initfunc(void trap_init(void))
 {
 	extern char except_vec0_r4000, except_vec0_r4600, except_vec0_r2300;
 	extern char except_vec1_generic, except_vec2_generic;
diff --git a/arch/mips/mips/Makefile b/arch/mips/mips/Makefile
new file mode 100644
index 000000000..b252e2fe4
--- /dev/null
+++ b/arch/mips/mips/Makefile
@@ -0,0 +1,39 @@
+#
+# PROM Entries for the big endian firmware used in the Mips Computer
+# System, Inc. machines.  Beware: some of the machines seem to be
+# different.
+#
+# Copyright (C) 1997 Ralf Baechle
+#
+prom-entries= reset exec restart reinit reboot autoboot open read write ioctl \
+	close getchar putchar showchar gets puts printf initproto protoenable \
+	protodisable getpkt putpkt orw_rmw orh_rmw orb_rmw andw_rmw andh_rmw \
+	andb_rmw flushcache clearcache setjmp longjmp bevutlb getenv setenv \
+	atob strcmp strlen strcpy strcat parser range argvize help dumpcmd \
+	setenvcmd unsetenvcmd printenvcmd bevexcept enablecmd disablecmd \
+	clearnofault notimplement nv_get nv_set
+asm-files=$(addsuffix .S,$(prom-entries))
+object-files=$(addsuffix .o,$(prom-entries))
+
+CC=mipsel-linux-gcc
+AR=mipsel-linux-ar
+CFLAGS=-O2 -mno-abicalls -fno-pic -G0 -Wall
+
+all:	libprom.a
+
+libprom.a: $(asm-files)
+	set -e;for i in $(prom-entries); do \
+		$(CC) $(CFLAGS) -c -o $$i.o $$i.S; \
+		$(AR) rcv libprom.a $$i.o; \
+	done
+
+$(asm-files): mkprom
+	set -e;for i in $(prom-entries); do \
+		mkprom $$i; \
+	done
+
+clean:
+	rm -f $(object-files)
+
+distclean:
+	rm -rf $(asm-files) libprom.a
diff --git a/arch/mips/mips/mkprom b/arch/mips/mips/mkprom
new file mode 100755
index 000000000..9e4cdf465
--- /dev/null
+++ b/arch/mips/mips/mkprom
@@ -0,0 +1,25 @@
+#
+# Generate PROM library stubs for the firmware used in the
+# Mips Computer System, Inc. machines.  Beware: some of the
+# machines seem to be different.
+#
+# Copyright (C) 1997 Ralf Baechle
+#
+fname=$1
+ucase=`echo $fname | tr 'a-z' 'A-Z'`
+cat << EOF | sed -e "s/@1@/$fname/" -e "s/@2@/$ucase/" >$fname.S
+/*
+ * WARNING: This file has been generated automatically.  Do not edit!
+ *
+ * Stub for the Mips firmware @1@() function.
+ */
+#include <asm/asm.h>
+#include <asm/regdef.h>
+#include <asm/mipsprom.h>
+
+LEAF(mips_prom_@1@)
+	lw	t0,__mips_prom_entry_offset
+	addu	t0,PROM_@2@
+	jr	t0
+	END(mips_prom_@1@)
+EOF
diff --git a/arch/mips/mm/Makefile b/arch/mips/mm/Makefile
index d12159903..1205b2bf3 100644
--- a/arch/mips/mm/Makefile
+++ b/arch/mips/mm/Makefile
@@ -9,6 +9,6 @@
 
 O_TARGET := mm.o
 O_OBJS	 := extable.o init.o fault.o r4xx0.o r2300.o r6000.o tfp.o \
-            andes.o loadmmu.o stack.o
+            andes.o loadmmu.o
 
 include $(TOPDIR)/Rules.make
diff --git a/arch/mips/mm/fault.c b/arch/mips/mm/fault.c
index 4fdd98155..3ad703b1f 100644
--- a/arch/mips/mm/fault.c
+++ b/arch/mips/mm/fault.c
@@ -70,7 +70,7 @@ good_area:
 		if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
 			goto bad_area;
 	}
-	handle_mm_fault(vma, address, writeaccess);
+	handle_mm_fault(tsk, vma, address, writeaccess);
 	up(&mm->mmap_sem);
 
 	goto out;
diff --git a/arch/mips/mm/r4xx0.c b/arch/mips/mm/r4xx0.c
index 5f1bcbfa6..f83f41df8 100644
--- a/arch/mips/mm/r4xx0.c
+++ b/arch/mips/mm/r4xx0.c
@@ -1,4 +1,4 @@
-/* $Id: r4xx0.c,v 1.19 1996/08/02 11:11:36 dm Exp $
+/*
  * r4xx0.c: R4000 processor variant specific MMU/Cache routines.
  *
  * Copyright (C) 1996 David S. Miller (dm@engr.sgi.com)
@@ -325,28 +325,6 @@ static inline void r4k_flush_cache_all_d32i32(void)
 	restore_flags(flags);
 }
 
-static inline struct vm_area_struct *
-find_mm_vma(struct mm_struct *mm, unsigned long addr)
-{
-	struct vm_area_struct * result = NULL;
-
-	if (mm) {
-		struct vm_area_struct * tree = mm->mmap_avl;
-		for (;;) {
-			if (tree == avl_empty)
-				break;
-			if (tree->vm_end > addr) {
-				result = tree;
-				if (tree->vm_start <= addr)
-					break;
-				tree = tree->vm_avl_left;
-			} else
-				tree = tree->vm_avl_right;
-		}
-	}
-	return result;
-}
-
 static void
 r4k_flush_cache_range_s16d16i16(struct mm_struct *mm,
                                 unsigned long start,
@@ -362,7 +340,7 @@ r4k_flush_cache_range_s16d16i16(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s16d16i16();
@@ -407,7 +385,7 @@ r4k_flush_cache_range_s32d16i16(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s32d16i16();
@@ -451,7 +429,7 @@ static void r4k_flush_cache_range_s64d16i16(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s64d16i16();
@@ -495,7 +473,7 @@ static void r4k_flush_cache_range_s128d16i16(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s128d16i16();
@@ -539,7 +517,7 @@ static void r4k_flush_cache_range_s16d32i32(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s16d32i32();
@@ -583,7 +561,7 @@ static void r4k_flush_cache_range_s32d32i32(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s32d32i32();
@@ -627,7 +605,7 @@ static void r4k_flush_cache_range_s64d32i32(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s64d32i32();
@@ -671,7 +649,7 @@ static void r4k_flush_cache_range_s128d32i32(struct mm_struct *mm,
 #ifdef DEBUG_CACHE
 	printk("crange[%d,%08lx,%08lx]", (int)mm->context, start, end);
 #endif
-	vma = find_mm_vma(mm, start);
+	vma = find_vma(mm, start);
 	if(vma) {
 		if(mm->context != current->mm->context) {
 			r4k_flush_cache_all_s128d32i32();
diff --git a/arch/mips/mm/stack.c b/arch/mips/mm/stack.c
deleted file mode 100644
index 89fb6dc64..000000000
--- a/arch/mips/mm/stack.c
+++ /dev/null
@@ -1,27 +0,0 @@
-/*
- * Kernel stack allocation/deallocation
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (C) 1996, 1997 by Ralf Baechle
- *
- * (This is _bad_ if the free page pool is fragmented ...)
- */
-#include <linux/sched.h>
-#include <linux/mm.h>
-#include <asm/processor.h>
-
-extern unsigned long alloc_kernel_stack(struct task_struct *tsk)
-{
-	unsigned long stack;
-	stack = __get_free_pages(GFP_KERNEL, 1, 0);
-
-	return stack;
-}
-
-extern void free_kernel_stack(unsigned long stack)
-{
-	free_pages(stack, 1);
-}
diff --git a/arch/mips/sgi/kernel/indy_int.c b/arch/mips/sgi/kernel/indy_int.c
index 19b744fdb..fe60c03c8 100644
--- a/arch/mips/sgi/kernel/indy_int.c
+++ b/arch/mips/sgi/kernel/indy_int.c
@@ -16,6 +16,8 @@
 #include <linux/timex.h>
 #include <linux/malloc.h>
 #include <linux/random.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
 
 #include <asm/bitops.h>
 #include <asm/bootinfo.h>
@@ -51,6 +53,7 @@ extern asmlinkage void indyIRQ(void);
 extern void rs_kgdb_hook(int);
 #endif
 
+unsigned int local_irq_count[NR_CPUS];
 unsigned long spurious_count = 0;
 
 /* Local IRQ's are layed out logically like this:
@@ -255,6 +258,15 @@ int get_irq_list(char *buf)
 	return len;
 }
 
+atomic_t __mips_bh_counter;
+
+#ifdef __SMP__
+#error Send superfluous SMP boxes to ralf@uni-koblenz.de
+#else
+#define irq_enter(cpu, irq)     (++local_irq_count[cpu])
+#define irq_exit(cpu, irq)      (--local_irq_count[cpu])
+#endif
+
 /*
  * do_IRQ handles IRQ's that have been installed without the
  * SA_INTERRUPT flag: it uses the full signal-handling return
@@ -264,8 +276,9 @@ int get_irq_list(char *buf)
  */
 asmlinkage void do_IRQ(int irq, struct pt_regs * regs)
 {
-	lock_kernel();
 	struct irqaction * action = *(irq + irq_action);
+
+	lock_kernel();
 	kstat.interrupts[irq]++;
 	printk("Got irq %d, press a key.", irq);
 	prom_getchar();
diff --git a/arch/mips/sgi/kernel/setup.c b/arch/mips/sgi/kernel/setup.c
index 58b7695b3..e0482e116 100644
--- a/arch/mips/sgi/kernel/setup.c
+++ b/arch/mips/sgi/kernel/setup.c
@@ -9,6 +9,7 @@
 #include <linux/kernel.h>
 #include <linux/sched.h>
 
+#include <asm/reboot.h>
 #include <asm/vector.h>
 #include <asm/sgialib.h>
 #include <asm/sgi.h>
diff --git a/arch/mips/sgi/prom/misc.c b/arch/mips/sgi/prom/misc.c
index 47051a1b3..53ee61cfe 100644
--- a/arch/mips/sgi/prom/misc.c
+++ b/arch/mips/sgi/prom/misc.c
@@ -1,9 +1,9 @@
-/* $Id: misc.c,v 1.3 1996/08/07 02:54:12 dm Exp $
+/*
  * misc.c: Miscellaneous ARCS PROM routines.
  *
  * Copyright (C) 1996 David S. Miller (dm@engr.sgi.com)
  */
-
+#include <linux/config.h>
 #include <linux/kernel.h>
 
 #include <asm/sgialib.h>
@@ -51,7 +51,9 @@ void prom_halt(void)
 {
 	shutoff_r4600_cache();
 	initialize_kbd();
+#if CONFIG_SCSI_SGIWD93
 	reset_wd33c93(sgiwd93_host);
+#endif
 	cli();
 	romvec->halt();
 }
@@ -60,7 +62,9 @@ void prom_powerdown(void)
 {
 	shutoff_r4600_cache();
 	initialize_kbd();
+#if CONFIG_SCSI_SGIWD93
 	reset_wd33c93(sgiwd93_host);
+#endif
 	cli();
 	romvec->pdown();
 }
@@ -70,7 +74,9 @@ void prom_restart(void)
 {
 	shutoff_r4600_cache();
 	initialize_kbd();
+#if CONFIG_SCSI_SGIWD93
 	reset_wd33c93(sgiwd93_host);
+#endif
 	cli();
 	romvec->restart();
 }
@@ -79,7 +85,9 @@ void prom_reboot(void)
 {
 	shutoff_r4600_cache();
 	initialize_kbd();
+#if CONFIG_SCSI_SGIWD93
 	reset_wd33c93(sgiwd93_host);
+#endif
 	cli();
 	romvec->reboot();
 }
@@ -88,7 +96,9 @@ void prom_imode(void)
 {
 	shutoff_r4600_cache();
 	initialize_kbd();
+#if CONFIG_SCSI_SGIWD93
 	reset_wd33c93(sgiwd93_host);
+#endif
 	cli();
 	romvec->imode();
 }
diff --git a/arch/mips/sni/.cvsignore b/arch/mips/sni/.cvsignore
new file mode 100644
index 000000000..4671378ae
--- /dev/null
+++ b/arch/mips/sni/.cvsignore
@@ -0,0 +1 @@
+.depend
diff --git a/arch/mips/sni/Makefile b/arch/mips/sni/Makefile
index f0bc3e600..9bd570ff0 100644
--- a/arch/mips/sni/Makefile
+++ b/arch/mips/sni/Makefile
@@ -13,7 +13,7 @@
 
 all: sni.o
 O_TARGET := sni.o
-O_OBJS	 := hw-access.o int-handler.o pci.o reset.o setup.o
+O_OBJS	 := hw-access.o int-handler.o io.o pci.o reset.o setup.o
 
 int-handler.o: int-handler.S
 
diff --git a/arch/mips/sni/int-handler.S b/arch/mips/sni/int-handler.S
index f60c82251..d051c8b13 100644
--- a/arch/mips/sni/int-handler.S
+++ b/arch/mips/sni/int-handler.S
@@ -1,7 +1,7 @@
 /*
  * SNI RM200 PCI specific interrupt handler code.
  *
- * Copyright (C) 1994, 1995, 1996 by Ralf Baechle
+ * Copyright (C) 1994 - 1997 by Ralf Baechle
  */
 #include <asm/asm.h>
 #include <linux/config.h>
@@ -150,30 +150,17 @@ poll_second:	li	a0,0x0f
 
 #endif /* CONFIG_PCNET32 */
 
-/*
- * FIXME:  This is definatly wrong but I'll have to do it this way
- *         'till I get more hardware info.
- * XXX: Apparently the NCR is attached to interrupt #2.
- */
 #ifdef CONFIG_SCSI_NCR53C8XX
 
 /*
- * FIXME: detect this address
- */
-#define NCR_BASE 0xb8000000
-
-/* Offsets from base I/O address. */
-#define NCR_INTF 0x14
-
-/*
  * ... check if we were interrupted by the NCR ...
  */
-3:		lb	t0,NCR_BASE+NCR_INTF
-		andi	t0,7
-		beqz	t0,3f				# no NCR interrupt?
+3:		lb	t0,PCIMT_CSITPEND
+		andi	t0,0x40
+		bnez	t0,3f				# bit 6 == 0 -> SCSI IRQ
 		nop					# delay slot
 		jal	do_fast_IRQ
-		li	a0,5				# delay slot
+		li	a0,PCIMT_IRQ_SCSI		# delay slot
 		j	return
 		nop					# delay slot
 
diff --git a/arch/mips/sni/io.c b/arch/mips/sni/io.c
new file mode 100644
index 000000000..f62fdc2d7
--- /dev/null
+++ b/arch/mips/sni/io.c
@@ -0,0 +1,172 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Low level I/O functions for SNI.
+ */
+#include <linux/string.h>
+#include <asm/mipsconfig.h>
+#include <asm/addrspace.h>
+#include <asm/system.h>
+#include <asm/spinlock.h>
+#include <asm/sni.h>
+
+unsigned char sni_map_isa_cache;
+
+#define unused __attribute__((unused))
+
+/*
+ * The PCIMT_CSMAPISA is shared by all processors; we need locking.
+ *
+ * XXX It's legal to use all the I/O memory access functions in interrupt
+ * code, so we need to use the _irq locking stuff which may result in
+ * significant IRQ latencies.
+ */
+static spinlock_t csmapisa_lock unused = SPIN_LOCK_UNLOCKED;
+
+/*
+ * Urgs...  We only can see a 16mb window of the 4gb EISA address space
+ * at PCIMT_EISA_BASE.  Maladia segmentitis ...
+ *
+ * XXX Check out if accessing PCIMT_CSMAPISA really is slow.
+ * For now assume so.
+ */
+static inline void update_isa_cache(unsigned long address)
+{
+	unsigned char upper;
+
+	upper = address >> 24;
+	if (sni_map_isa_cache != upper) {
+		sni_map_isa_cache = upper;
+		*(volatile unsigned char *)PCIMT_CSMAPISA = ~upper;
+	}
+}
+
+static unsigned char sni_readb(unsigned long addr)
+{
+	unsigned char res;
+
+	spin_lock_irq(&csmapisa_lock);
+	update_isa_cache(addr);
+	addr &= 0xffffff;
+	res = *(volatile unsigned char *) (PCIMT_EISA_BASE + addr);
+	spin_unlock_irq(&csmapisa_lock);
+
+	return res;
+}
+
+static unsigned short sni_readw(unsigned long addr)
+{
+	unsigned short res;
+
+	spin_lock_irq(&csmapisa_lock);
+	update_isa_cache(addr);
+	addr &= 0xffffff;
+	res = *(volatile unsigned char *) (PCIMT_EISA_BASE + addr);
+	spin_unlock_irq(&csmapisa_lock);
+
+	return res;
+}
+
+static unsigned int sni_readl(unsigned long addr)
+{
+	unsigned int res;
+
+	spin_lock_irq(&csmapisa_lock);
+	update_isa_cache(addr);
+	addr &= 0xffffff;
+	res = *(volatile unsigned char *) (PCIMT_EISA_BASE + addr);
+	spin_unlock_irq(&csmapisa_lock);
+
+	return res;
+}
+
+static void sni_writeb(unsigned char val, unsigned long addr)
+{
+	spin_lock_irq(&csmapisa_lock);
+	update_isa_cache(addr);
+	addr &= 0xffffff;
+	*(volatile unsigned char *) (PCIMT_EISA_BASE + addr) = val;
+	spin_unlock_irq(&csmapisa_lock);
+}
+
+static void sni_writew(unsigned short val, unsigned long addr)
+{
+	spin_lock_irq(&csmapisa_lock);
+	update_isa_cache(addr);
+	addr &= 0xffffff;
+	*(volatile unsigned char *) (PCIMT_EISA_BASE + addr) = val;
+	spin_unlock_irq(&csmapisa_lock);
+}
+
+static void sni_writel(unsigned int val, unsigned long addr)
+{
+	spin_lock_irq(&csmapisa_lock);
+	update_isa_cache(addr);
+	addr &= 0xffffff;
+	*(volatile unsigned char *) (PCIMT_EISA_BASE + addr) = val;
+	spin_unlock_irq(&csmapisa_lock);
+}
+
+static void sni_memset_io(unsigned long addr, int val, unsigned long len)
+{
+	unsigned long waddr;
+
+	waddr = PCIMT_EISA_BASE | (addr & 0xffffff);
+	spin_lock_irq(&csmapisa_lock);
+	while(len) {
+		unsigned long fraglen;
+
+		fraglen = (~addr + 1) & 0xffffff;
+		fraglen = (fraglen < len) ? fraglen : len;
+		update_isa_cache(addr);
+		memset((char *)waddr, val, fraglen);
+		addr += fraglen;
+		waddr = waddr + fraglen - 0x1000000;
+		len -= fraglen;
+	}
+	spin_unlock_irq(&csmapisa_lock);
+}
+
+static void sni_memcpy_fromio(unsigned long to, unsigned long from, unsigned long len)
+{
+	unsigned long waddr;
+
+	waddr = PCIMT_EISA_BASE | (from & 0xffffff);
+	spin_lock_irq(&csmapisa_lock);
+	while(len) {
+		unsigned long fraglen;
+
+		fraglen = (~from + 1) & 0xffffff;
+		fraglen = (fraglen < len) ? fraglen : len;
+		update_isa_cache(from);
+		memcpy((void *)to, (void *)waddr, fraglen);
+		to += fraglen;
+		from += fraglen;
+		waddr = waddr + fraglen - 0x1000000;
+		len -= fraglen;
+	}
+	spin_unlock_irq(&csmapisa_lock);
+}
+
+static void sni_memcpy_toio(unsigned long to, unsigned long from, unsigned long len)
+{
+	unsigned long waddr;
+
+	waddr = PCIMT_EISA_BASE | (to & 0xffffff);
+	spin_lock_irq(&csmapisa_lock);
+	while(len) {
+		unsigned long fraglen;
+
+		fraglen = (~to + 1) & 0xffffff;
+		fraglen = (fraglen < len) ? fraglen : len;
+		update_isa_cache(to);
+		memcpy((char *)to + PCIMT_EISA_BASE, (void *)from, fraglen);
+		to += fraglen;
+		from += fraglen;
+		waddr = waddr + fraglen - 0x1000000;
+		len -= fraglen;
+	}
+	spin_unlock_irq(&csmapisa_lock);
+}
diff --git a/arch/mips/sni/pci.c b/arch/mips/sni/pci.c
index 06edb268f..917d07a81 100644
--- a/arch/mips/sni/pci.c
+++ b/arch/mips/sni/pci.c
@@ -7,6 +7,7 @@
  */
 #include <linux/config.h>
 #include <linux/bios32.h>
+#include <linux/init.h>
 #include <linux/pci.h>
 #include <linux/types.h>
 #include <asm/byteorder.h>
@@ -18,18 +19,26 @@
 extern inline u32 mkaddr(unsigned char bus, unsigned char dev_fn,
                          unsigned char where)
 {
-	return (((bus    & 0xff) << 0x10) |
-	        ((dev_fn & 0xff) << 0x08) |
-	        ((where  & 0xfc)));
+	return ((bus    & 0xff) << 0x10) |
+	       ((dev_fn & 0xff) << 0x08) |
+	       (where  & 0xfc);
 }
 
 static unsigned long sni_rm200_pcibios_fixup (unsigned long memory_start,
                                               unsigned long memory_end)
 {
-	/* I guess it's ok to do exactly nothing.  */
+	/*
+	 * TODO: Fix PCI_INTERRUPT_LINE register for onboard cards.
+	 * Take care of RM300 revision D boards for where the network
+	 * slot became an ordinary PCI slot.
+	 */
 	return memory_start;
 }
 
+/*
+ * We can't address 8 and 16 bit words directly.  Instead we have to
+ * read/write a 32bit word and mask/modify the data we actually want.
+ */
 static int sni_rm200_pcibios_read_config_byte (unsigned char bus,
                                                unsigned char dev_fn,
                                                unsigned char where,
@@ -39,7 +48,7 @@ static int sni_rm200_pcibios_read_config_byte (unsigned char bus,
 
 	*(volatile u32 *)PCIMT_CONFIG_ADDRESS = mkaddr(bus, dev_fn, where);
 	res = *(volatile u32 *)PCIMT_CONFIG_DATA;
-	res = le32_to_cpu(res);
+	res = (le32_to_cpu(res) >> ((where & 3) << 3)) & 0xff;
 	*val = res;
 
 	return PCIBIOS_SUCCESSFUL;
@@ -56,7 +65,7 @@ static int sni_rm200_pcibios_read_config_word (unsigned char bus,
 		return PCIBIOS_BAD_REGISTER_NUMBER;
 	*(volatile u32 *)PCIMT_CONFIG_ADDRESS = mkaddr(bus, dev_fn, where);
 	res = *(volatile u32 *)PCIMT_CONFIG_DATA;
-	res = le32_to_cpu(res);
+	res = (le32_to_cpu(res) >> ((where & 3) << 3)) & 0xffff;
 	*val = res;
 
 	return PCIBIOS_SUCCESSFUL;
@@ -110,7 +119,7 @@ static int sni_rm200_pcibios_write_config_dword (unsigned char bus,
 	return PCIBIOS_SUCCESSFUL;
 }
 
-unsigned long sni_rm200_pcibios_init(unsigned long memory_start, unsigned long memory_end)
+__initfunc(unsigned long sni_rm200_pcibios_init(unsigned long memory_start, unsigned long memory_end))
 {
 	_pcibios_fixup = sni_rm200_pcibios_fixup;
 	_pcibios_read_config_byte = sni_rm200_pcibios_read_config_byte;
diff --git a/arch/mips/sni/setup.c b/arch/mips/sni/setup.c
index 1a7a1387d..7fa76a490 100644
--- a/arch/mips/sni/setup.c
+++ b/arch/mips/sni/setup.c
@@ -5,11 +5,12 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 1996 by Ralf Baechle
+ * Copyright (C) 1996, 1997 by Ralf Baechle
  */
 #include <asm/ptrace.h>
 #include <linux/ioport.h>
 #include <linux/sched.h>
+#include <linux/init.h>
 #include <linux/interrupt.h>
 #include <linux/timex.h>
 #include <linux/pci.h>
@@ -40,8 +41,7 @@ extern void sni_machine_restart(char *command);
 extern void sni_machine_halt(void);
 extern void sni_machine_power_off(void);
 
-static void
-sni_irq_setup(void)
+__initfunc(static void sni_irq_setup(void))
 {
 	set_except_vector(0, sni_rm200_pci_handle_int);
 	request_region(0x20,0x20, "pic1");
@@ -57,7 +57,7 @@ sni_irq_setup(void)
 
 void (*board_time_init)(struct irqaction *irq);
 
-static void sni_rm200_pci_time_init(struct irqaction *irq)
+__initfunc(static void sni_rm200_pci_time_init(struct irqaction *irq))
 {
 	/* set the clock to 100 Hz */
 	outb_p(0x34,0x43);		/* binary, mode 2, LSB/MSB, ch 0 */
@@ -67,8 +67,9 @@ static void sni_rm200_pci_time_init(struct irqaction *irq)
 }
 
 unsigned char aux_device_present;
-unsigned long sni_rm200_pcibios_init (unsigned long memory_start,
-                                      unsigned long memory_end);
+extern unsigned long sni_rm200_pcibios_init (unsigned long memory_start,
+                                             unsigned long memory_end);
+extern unsigned char sni_map_isa_cache;
 
 /*
  * A bit more gossip about the iron we're running on ...
@@ -92,8 +93,7 @@ static inline void sni_pcimt_detect(void)
 	printk("%s.\n", boardtype);
 }
 	
-void
-sni_rm200_pci_setup(void)
+__initfunc(void sni_rm200_pci_setup(void))
 {
 	tag *atag;
 
@@ -127,7 +127,14 @@ sni_rm200_pci_setup(void)
 	fd_cacheflush = sni_fd_cacheflush;	// Will go away
 	feature = &sni_rm200_pci_feature;
 	port_base = SNI_PORT_BASE;
+
+	/*
+	 * Setup (E)ISA I/O memory access stuff
+	 */
 	isa_slot_offset = 0xb0000000;
+	// sni_map_isa_cache = 0;
+	EISA_bus = 1;
+
 	request_region(0x00,0x20,"dma1");
 	request_region(0x40,0x20,"timer");
 	/* XXX FIXME: CONFIG_RTC */
@@ -140,8 +147,6 @@ sni_rm200_pci_setup(void)
 	_machine_halt = sni_machine_halt;
 	_machine_power_off = sni_machine_power_off;
 
-	if (mips_machtype == MACH_SNI_RM200_PCI)
-		EISA_bus = 1;
 	aux_device_present = 0xaa;
 
 	/*
diff --git a/arch/mips/tools/.cvsignore b/arch/mips/tools/.cvsignore
new file mode 100644
index 000000000..2a1642034
--- /dev/null
+++ b/arch/mips/tools/.cvsignore
@@ -0,0 +1 @@
+.depend offset.s offset.h
diff --git a/arch/mips/tools/offset.c b/arch/mips/tools/offset.c
index 3bb46de6e..a8660dcf6 100644
--- a/arch/mips/tools/offset.c
+++ b/arch/mips/tools/offset.c
@@ -77,12 +77,11 @@ void output_task_defines(void)
 {
 	text("/* MIPS task_struct offsets. */");
 	offset("#define TASK_STATE         ", struct task_struct, state);
+	offset("#define TASK_COUNTER       ", struct task_struct, counter);
 	offset("#define TASK_PRIORITY      ", struct task_struct, priority);
 	offset("#define TASK_SIGNAL        ", struct task_struct, signal);
 	offset("#define TASK_BLOCKED       ", struct task_struct, blocked);
 	offset("#define TASK_FLAGS         ", struct task_struct, flags);
-	offset("#define TASK_SAVED_KSTACK  ", struct task_struct, saved_kernel_stack);
-	offset("#define TASK_KSTACK_PG     ", struct task_struct, kernel_stack_page);
 	offset("#define TASK_MM            ", struct task_struct, mm);
 	linefeed;
 }
diff --git a/arch/mips/tools/offset.h b/arch/mips/tools/offset.h
deleted file mode 100644
index c5e6f37fa..000000000
--- a/arch/mips/tools/offset.h
+++ /dev/null
@@ -1,89 +0,0 @@
-/* DO NOT TOUCH, AUTOGENERATED BY OFFSET.C */
-
-#ifndef _MIPS_OFFSET_H
-#define _MIPS_OFFSET_H
-
-/* MIPS pt_regs offsets. */
-#define PT_R0     24
-#define PT_R1     28
-#define PT_R2     32
-#define PT_R3     36
-#define PT_R4     40
-#define PT_R5     44
-#define PT_R6     48
-#define PT_R7     52
-#define PT_R8     56
-#define PT_R9     60
-#define PT_R10    64
-#define PT_R11    68
-#define PT_R12    72
-#define PT_R13    76
-#define PT_R14    80
-#define PT_R15    84
-#define PT_R16    88
-#define PT_R17    92
-#define PT_R18    96
-#define PT_R19    100
-#define PT_R20    104
-#define PT_R21    108
-#define PT_R22    112
-#define PT_R23    116
-#define PT_R24    120
-#define PT_R25    124
-#define PT_R26    128
-#define PT_R27    132
-#define PT_R28    136
-#define PT_R29    140
-#define PT_R30    144
-#define PT_R31    148
-#define PT_LO     152
-#define PT_HI     156
-#define PT_OR2    160
-#define PT_OR7    164
-#define PT_EPC    168
-#define PT_BVADDR 172
-#define PT_STATUS 176
-#define PT_CAUSE  180
-#define PT_SIZE   184
-
-/* MIPS task_struct offsets. */
-#define TASK_STATE         0
-#define TASK_PRIORITY      8
-#define TASK_SIGNAL        12
-#define TASK_BLOCKED       16
-#define TASK_FLAGS         20
-#define TASK_SAVED_KSTACK  84
-#define TASK_KSTACK_PG     88
-#define TASK_MM            912
-
-/* MIPS specific thread_struct offsets. */
-#define THREAD_REG16   544
-#define THREAD_REG17   548
-#define THREAD_REG18   552
-#define THREAD_REG19   556
-#define THREAD_REG20   560
-#define THREAD_REG21   564
-#define THREAD_REG22   568
-#define THREAD_REG23   572
-#define THREAD_REG28   576
-#define THREAD_REG29   580
-#define THREAD_REG30   584
-#define THREAD_REG31   588
-#define THREAD_STATUS  592
-#define THREAD_FPU     600
-#define THREAD_BVADDR  864
-#define THREAD_ECODE   868
-#define THREAD_TRAPNO  872
-#define THREAD_KSP     876
-#define THREAD_PGDIR   880
-#define THREAD_MFLAGS  884
-#define THREAD_CURDS   888
-#define THREAD_TRAMP   892
-#define THREAD_OLDCTX  896
-
-/* Linux mm_struct offsets. */
-#define MM_COUNT      0
-#define MM_PGD        4
-#define MM_CONTEXT    8
-
-#endif /* !(_MIPS_OFFSET_H) */
diff --git a/arch/ppc/kernel/misc.S b/arch/ppc/kernel/misc.S
index e137e1171..c2a2989ec 100644
--- a/arch/ppc/kernel/misc.S
+++ b/arch/ppc/kernel/misc.S
@@ -855,5 +855,6 @@ sys_call_table:
 	.long sys_mremap
 	.long SYMBOL_NAME(sys_setresuid)
 	.long SYMBOL_NAME(sys_getresuid)
-	.space (NR_syscalls-165)*4
+	.long SYMBOL_NAME(sys_nfsservctl)
+	.space (NR_syscalls-166)*4
 
diff --git a/arch/ppc/kernel/process.c b/arch/ppc/kernel/process.c
index 8dc0e61ca..ee21c5cbe 100644
--- a/arch/ppc/kernel/process.c
+++ b/arch/ppc/kernel/process.c
@@ -29,6 +29,24 @@
 #include <asm/io.h>
 
 #include <asm/ppc_machine.h>
+
+/*
+ * Initial task structure. Make this a per-architecture thing,
+ * because different architectures tend to have different
+ * alignment requirements and potentially different initial
+ * setup.
+ */
+static unsigned long init_kernel_stack[1024] = { STACK_MAGIC, };
+unsigned long init_user_stack[1024] = { STACK_MAGIC, };
+static struct vm_area_struct init_mmap = INIT_MMAP;
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS;
+
+struct mm_struct init_mm = INIT_MM;
+struct task_struct init_task = INIT_TASK;
+
+
 int dump_fpu(void);
 void hard_reset_now(void);
 void switch_to(struct task_struct *, struct task_struct *);
diff --git a/arch/ppc/kernel/ptrace.c b/arch/ppc/kernel/ptrace.c
index 2d960c8cd..dc8a95302 100644
--- a/arch/ppc/kernel/ptrace.c
+++ b/arch/ppc/kernel/ptrace.c
@@ -42,18 +42,6 @@
 static int regoff[] = {
 };
 
-/* change a pid into a task struct. */
-static inline struct task_struct * get_task(int pid)
-{
-	int i;
-
-	for (i = 1; i < NR_TASKS; i++) {
-		if (task[i] != NULL && (task[i]->pid == pid))
-			return task[i];
-	}
-	return NULL;
-}
-
 /*
  * Get contents of register REGNO in task TASK.
  */
@@ -394,7 +382,7 @@ asmlinkage int sys_ptrace(long request, long pid, long addr, long data)
 	if (pid == 1)		/* you may not mess with init */
 		goto out;
 	ret = -ESRCH;
-	if (!(child = get_task(pid)))
+	if (!(child = find_task_by_pid(pid)))
 		goto out;
 	ret = -EPERM;
 	if (request == PTRACE_ATTACH) {
diff --git a/arch/ppc/mm/init.c b/arch/ppc/mm/init.c
index 6a6ffdb89..ee2381dba 100644
--- a/arch/ppc/mm/init.c
+++ b/arch/ppc/mm/init.c
@@ -58,69 +58,62 @@ pte_t __bad_page(void)
 
 void show_mem(void)
 {
-  unsigned long i,free = 0,total = 0,reserved = 0;
-  unsigned long shared = 0;
-  PTE *ptr;
-  unsigned long full = 0, overflow = 0;
-  unsigned int ti;
-
-  printk("Mem-info:\n");
-  show_free_areas();
-  printk("Free swap:       %6dkB\n",nr_swap_pages<<(PAGE_SHIFT-10));
-  i = MAP_NR(high_memory);
-  while (i-- > 0) {
-    total++;
-    if (PageReserved(mem_map+i))
-      reserved++;
-    else if (!atomic_read(&mem_map[i].count))
-      free++;
-    else
-      shared += atomic_read(&mem_map[i].count) - 1;
-  }
-  printk("%lu pages of RAM\n",total);
-  printk("%lu free pages\n",free);
-  printk("%lu reserved pages\n",reserved);
-  printk("%lu pages shared\n",shared);
-  show_buffers();
+	struct task_struct *p;
+	unsigned long i,free = 0,total = 0,reserved = 0;
+	unsigned long shared = 0;
+	PTE *ptr;
+	unsigned long full = 0, overflow = 0;
+	unsigned int ti;
+
+	printk("Mem-info:\n");
+	show_free_areas();
+	printk("Free swap:       %6dkB\n",nr_swap_pages<<(PAGE_SHIFT-10));
+	i = MAP_NR(high_memory);
+	while (i-- > 0) {
+		total++;
+		if (PageReserved(mem_map+i))
+			reserved++;
+		else if (!atomic_read(&mem_map[i].count))
+			free++;
+		else
+			shared += atomic_read(&mem_map[i].count) - 1;
+	}
+	printk("%lu pages of RAM\n",total);
+	printk("%lu free pages\n",free);
+	printk("%lu reserved pages\n",reserved);
+	printk("%lu pages shared\n",shared);
+	show_buffers();
 #ifdef CONFIG_NET
-  show_net_buffers();
+	show_net_buffers();
 #endif
 #ifdef HASHSTATS
-  printk("Hash Hits %u entries (buckets)\n",(Hash_size/sizeof(struct _PTE))/8);
-  for ( i = 0; i < (Hash_size/sizeof(struct _PTE))/8; i++ )
-  {
-    if ( hashhits[i] >= 20 )
-    {
-      printk("[%lu] \t %lu\n", i,hashhits[i]);
-    }
-  }
+	printk("Hash Hits %u entries (buckets)\n",(Hash_size/sizeof(struct _PTE))/8);
+	for ( i = 0; i < (Hash_size/sizeof(struct _PTE))/8; i++ ) {
+		if ( hashhits[i] >= 20 )
+			printk("[%lu] \t %lu\n", i,hashhits[i]);
+	}
 #endif
   
-  for ( ptr = Hash ; ptr <= Hash+Hash_size ; ptr++)
-  {
-    if (ptr->v)
-    {
-      full++;
-      if (ptr->h == 1)
-	overflow++;
-    }
-  }
-  printk("Hash Table: %dkB Buckets: %dk PTEs: %d/%d (%%%d full) %d overflowed\n",
-	 Hash_size>>10,	(Hash_size/(sizeof(PTE)*8)) >> 10,
-	 full,Hash_size/sizeof(PTE),
-	 (full*100)/(Hash_size/sizeof(PTE)),
-	 overflow);
-  printk(" Task  context    vsid0\n");
-  for ( ti = 0; ti < NR_TASKS ; ti++ );
-  {
-    if ( task[ti] )
-    {
-      printk("%5d %8x %8x\n",
-	     task[ti]->pid,task[ti]->mm->context,
-	     ((SEGREG *)task[ti]->tss.segs)[0].vsid);
-    }
-  }
-      
+	for ( ptr = Hash ; ptr <= Hash+Hash_size ; ptr++) {
+		if (ptr->v) {
+			full++;
+			if (ptr->h == 1)
+				overflow++;
+		}
+	}
+	printk("Hash Table: %dkB Buckets: %dk PTEs: %d/%d (%%%d full) %d overflowed\n",
+	       Hash_size>>10,	(Hash_size/(sizeof(PTE)*8)) >> 10,
+	       full,Hash_size/sizeof(PTE),
+	       (full*100)/(Hash_size/sizeof(PTE)),
+	       overflow);
+	printk(" Task  context    vsid0\n");
+	read_lock(&tasklist_lock);
+	for_each_task(p) {
+		printk("%5d %8x %8x\n",
+		       p->pid,p->mm->context,
+		       ((SEGREG *)p->tss.segs)[0].vsid);
+	}
+	read_unlock(&tasklist_lock);
 }
 
 extern unsigned long free_area_init(unsigned long, unsigned long);
diff --git a/arch/sparc/Makefile b/arch/sparc/Makefile
index 587cfc616..92c73de32 100644
--- a/arch/sparc/Makefile
+++ b/arch/sparc/Makefile
@@ -1,4 +1,4 @@
-# $Id: Makefile,v 1.27 1997/04/07 06:54:08 davem Exp $
+# $Id: Makefile,v 1.28 1997/05/01 01:41:21 davem Exp $
 # sparc/Makefile
 #
 # Makefile for the architecture dependent flags and dependencies on the
@@ -21,7 +21,7 @@ CFLAGS := $(CFLAGS) -pipe -fcall-used-g5 -fcall-used-g7
 #LINKFLAGS = -N -Ttext 0xf0004000
 LINKFLAGS = -T arch/sparc/vmlinux.lds
 
-HEAD := arch/sparc/kernel/head.o
+HEAD := arch/sparc/kernel/head.o arch/sparc/kernel/init_task.o
 
 SUBDIRS := $(SUBDIRS) arch/sparc/kernel arch/sparc/lib arch/sparc/mm \
 	arch/sparc/prom
diff --git a/arch/sparc/ap1000/bnet.c b/arch/sparc/ap1000/bnet.c
index 9305b0a55..974501d79 100644
--- a/arch/sparc/ap1000/bnet.c
+++ b/arch/sparc/ap1000/bnet.c
@@ -872,12 +872,12 @@ static void reply_kill(struct cap_request *req)
   len = req->size - sizeof(*req);
   if (len == 0) {
     int pid = req->data[1];
-    for_each_task(p) 
-      if (p->pid == pid) {
-	send_sig(sig,p,1);
-	return;
-      }
-    printk("cell %d: no task with pid %d\n",mpp_cid(),pid);
+    p = find_task_by_pid(pid);
+
+    if(p)
+	    send_sig(sig, p, 1);
+    else
+	    printk("cell %d: no task with pid %d\n",mpp_cid(),pid);
     return;
   }
 
@@ -889,9 +889,11 @@ static void reply_kill(struct cap_request *req)
   read_bif(name,len);
   name[len] = 0;
 
+  read_lock(&tasklist_lock);
   for_each_task(p) 
     if (strcmp(name,p->comm) == 0)
       send_sig(sig,p,1);
+  read_unlock(&tasklist_lock);
 }
 
 
diff --git a/arch/sparc/ap1000/hw.c b/arch/sparc/ap1000/hw.c
index c36150c4e..aabe15a3b 100644
--- a/arch/sparc/ap1000/hw.c
+++ b/arch/sparc/ap1000/hw.c
@@ -86,14 +86,21 @@ static void show_task(struct task_struct *t)
 static void show_ptasks(void)
 {
 	extern struct task_struct *task[];
+	struct task_struct *p;
 	int i;
 	int count=0;
 
-	for (i=MPP_TASK_BASE;i<NR_TASKS;i++)
-		if (task[i]) {
-			show_task(task[i]);
+	read_lock(&tasklist_lock);
+	for_each_task(p) {
+		struct task_struct **tp = p->tarray_ptr;
+
+		if(tp >= &task[MPP_TASK_BASE]) {
+			show_task(p);
 			count++;
 		}
+	}
+	read_unlock(&tasklist_lock);
+
 	if (count == 0)
 		printk("no parallel tasks on cell %d\n",mpp_cid());
 }
@@ -101,14 +108,18 @@ static void show_ptasks(void)
 static void show_utasks(void)
 {
 	extern struct task_struct *task[];
+	struct task_struct *p;
 	int i;
 	int count=0;
 
-	for (i=0;i<NR_TASKS;i++)
-		if (task[i] && task[i]->uid > 1)  {
+	read_lock(&tasklist_lock);
+	for_each_task(p) {
+		if(p->uid > 1) {
 			show_task(task[i]);
 			count++;
 		}
+	}
+	read_unlock(&tasklist_lock);
 
 	if (count == 0)
 		printk("no user tasks on cell %d\n",mpp_cid());
@@ -118,15 +129,19 @@ static void show_utasks(void)
 static void show_otasks(void)
 {
 	extern struct task_struct *task[];
+	struct task_struct *p;
 	int i;
 	int count=0;
 	extern int ap_current_uid;
 
-	for (i=0;i<NR_TASKS;i++)
-		if (task[i] && task[i]->uid == ap_current_uid)  {
+	read_lock(&tasklist_lock);
+	for_each_task(p) {
+		if(p->uid == ap_current_uid) {
 			show_task(task[i]);
 			count++;
 		}
+	}
+	read_unlock(&tasklist_lock);
 
 	if (count == 0)
 		printk("no tasks on cell %d\n",mpp_cid());
diff --git a/arch/sparc/ap1000/mpp.c b/arch/sparc/ap1000/mpp.c
index 021c1a2e5..84ef7f28e 100644
--- a/arch/sparc/ap1000/mpp.c
+++ b/arch/sparc/ap1000/mpp.c
@@ -66,6 +66,7 @@ int mpp_weight(struct task_struct *tsk)
 
 	if (block_parallel_tasks) return -1000;
 
+	/* XXX task[] fixme */
 	if (last_task && last_task != tsk->taskid && task[last_task] &&
 	    !msc_switch_ok()) return -1000;
 
diff --git a/arch/sparc/kernel/Makefile b/arch/sparc/kernel/Makefile
index f73a52fb3..d3307e463 100644
--- a/arch/sparc/kernel/Makefile
+++ b/arch/sparc/kernel/Makefile
@@ -1,4 +1,4 @@
-# $Id: Makefile,v 1.39 1997/04/01 02:21:44 davem Exp $
+# $Id: Makefile,v 1.40 1997/05/01 01:40:36 davem Exp $
 # Makefile for the linux kernel.
 #
 # Note! Dependencies are done automagically by 'make dep', which also
@@ -28,7 +28,7 @@ else
 CHECKASM_CC	=	$(CC)
 endif
 
-all: kernel.o head.o
+all: kernel.o head.o init_task.o
 
 O_TARGET := kernel.o
 IRQ_OBJS := irq.o sun4m_irq.o sun4c_irq.o
diff --git a/arch/sparc/kernel/etrap.S b/arch/sparc/kernel/etrap.S
index b93094919..fc6c4cead 100644
--- a/arch/sparc/kernel/etrap.S
+++ b/arch/sparc/kernel/etrap.S
@@ -1,4 +1,4 @@
-/* $Id: etrap.S,v 1.23 1997/03/04 16:26:25 jj Exp $
+/* $Id: etrap.S,v 1.26 1997/05/01 08:53:32 davem Exp $
  * etrap.S: Sparc trap window preparation for entry into the
  *          Linux kernel.
  *
@@ -129,15 +129,23 @@ tsetup_patch2:
 trap_setup_from_user:
 	/* We can't use %curptr yet. */
 	LOAD_CURRENT(t_kstack, t_twinmask)
+
+	mov	1, %t_twinmask
+	sll	%t_twinmask, (PAGE_SHIFT + 1), %t_twinmask
+	sub	%t_twinmask, (TRACEREG_SZ + REGWIN_SZ), %t_twinmask
+	add	%t_kstack, %t_twinmask, %t_kstack
+
 	mov	1, %t_twinmask
-	ld	[%t_kstack + AOFF_task_saved_kernel_stack], %t_kstack
 	sll	%t_twinmask, %t_psr, %t_twinmask ! t_twinmask = (1 << psr)
 
 	/* Build pt_regs frame. */
 	STORE_PT_ALL(t_kstack, t_psr, t_pc, t_npc, g2)
 
 	/* Clear current->tss.w_saved */
-	LOAD_CURRENT(curptr, g1)
+	mov	1, %curptr
+	sll	%curptr, (PAGE_SHIFT + 1), %curptr
+	sub	%curptr, (TRACEREG_SZ + REGWIN_SZ), %curptr
+	sub	%t_kstack, %curptr, %curptr
 	st	%g0, [%curptr + AOFF_task_tss + AOFF_thread_w_saved]
 
 	/* See if we are in the trap window. */
diff --git a/arch/sparc/kernel/head.S b/arch/sparc/kernel/head.S
index eea3528bf..90965e26e 100644
--- a/arch/sparc/kernel/head.S
+++ b/arch/sparc/kernel/head.S
@@ -1,4 +1,4 @@
-/* $Id: head.S,v 1.81 1997/04/16 07:15:48 davem Exp $
+/* $Id: head.S,v 1.82 1997/05/01 01:40:38 davem Exp $
  * head.S: The initial boot code for the Sparc port of Linux.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -393,12 +393,6 @@ C_LABEL(trapbase_cpu3):
 	BAD_TRAP(0xf7) BAD_TRAP(0xf8) BAD_TRAP(0xf9) BAD_TRAP(0xfa) BAD_TRAP(0xfb)
 	BAD_TRAP(0xfc) BAD_TRAP(0xfd) BAD_TRAP(0xfe) BAD_TRAP(0xff)
 
-	.globl	C_LABEL(cpu0_stack), C_LABEL(cpu1_stack), C_LABEL(cpu2_stack)
-	.globl	C_LABEL(cpu3_stack)
-C_LABEL(cpu0_stack):			.skip 0x2000
-C_LABEL(cpu1_stack):			.skip 0x2000
-C_LABEL(cpu2_stack):			.skip 0x2000
-C_LABEL(cpu3_stack):			.skip 0x2000
 #endif
 	.skip 4096
 
@@ -742,6 +736,9 @@ go_to_highmem:
 		jmpl	%g1, %g0
 		 nop
 
+	/* This is to align init_task_union properly, be careful. -DaveM */
+	.align	8192
+
 /* The code above should be at beginning and we have to take care about
  * short jumps, as branching to .text.init section from .text is usually
  * impossible */
@@ -986,12 +983,10 @@ sun4c_continue_boot:
 		/* Initialize the umask value for init_task just in case.
 		 * But first make current_set[0] point to something useful.
 		 */
-		set	C_LABEL(init_task), %g6
+		set	C_LABEL(init_task_union), %g6
 		set	C_LABEL(current_set), %g2
 		st	%g6, [%g2]
 
-		set	C_LABEL(bootup_kernel_stack), %g3
-		st	%g3, [%g6 + AOFF_task_kernel_stack_page]
 		st	%g0, [%g6 + AOFF_task_tss + AOFF_thread_uwinmask]
 
 /* Compute NWINDOWS and stash it away. Now uses %wim trick explained
diff --git a/arch/sparc/kernel/init_task.c b/arch/sparc/kernel/init_task.c
new file mode 100644
index 000000000..d0fc09346
--- /dev/null
+++ b/arch/sparc/kernel/init_task.c
@@ -0,0 +1,18 @@
+#include <linux/mm.h>
+#include <linux/sched.h>
+
+#include <asm/pgtable.h>
+#include <asm/uaccess.h>
+
+static struct vm_area_struct init_mmap = INIT_MMAP;
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS;
+struct mm_struct init_mm = INIT_MM;
+
+/* .text section in head.S is aligned at 2 page boundry and this gets linked
+ * right after that so that the init_task_union is aligned properly as well.
+ * We really don't need this special alignment like the Intel does, but
+ * I do it anyways for completeness.
+ */
+union task_union init_task_union __attribute__((__section__(".text"))) = { INIT_TASK };
diff --git a/arch/sparc/kernel/irq.c b/arch/sparc/kernel/irq.c
index 70042857f..b22f008a1 100644
--- a/arch/sparc/kernel/irq.c
+++ b/arch/sparc/kernel/irq.c
@@ -1,4 +1,4 @@
-/*  $Id: irq.c,v 1.72 1997/04/20 11:41:26 ecd Exp $
+/*  $Id: irq.c,v 1.75 1997/05/08 20:57:37 davem Exp $
  *  arch/sparc/kernel/irq.c:  Interrupt request handling routines. On the
  *                            Sparc the IRQ's are basically 'cast in stone'
  *                            and you are supposed to probe the prom's device
@@ -313,8 +313,180 @@ spinlock_t global_bh_lock = SPIN_LOCK_UNLOCKED;
 /* Global IRQ locking depth. */
 atomic_t global_irq_count = ATOMIC_INIT(0);
 
+#ifdef DEBUG_IRQLOCK
+
+static unsigned long previous_irqholder;
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if (!--stuck) {printk("wait_on_irq CPU#%d stuck at %08lx, waiting for %08lx (local=%d, global=%d)\n", cpu, where, previous_irqholder, local_count, atomic_read(&global_irq_count)); stuck = INIT_STUCK; }
+
+static inline void wait_on_irq(int cpu, unsigned long where)
+{
+	int stuck = INIT_STUCK;
+	int local_count = local_irq_count[cpu];
+
+	/* Are we the only one in an interrupt context? */
+	while (local_count != atomic_read(&global_irq_count)) {
+		/*
+		 * No such luck. Now we need to release the lock,
+		 * _and_ release our interrupt context, because
+		 * otherwise we'd have dead-locks and live-locks
+		 * and other fun things.
+		 */
+		atomic_sub(local_count, &global_irq_count);
+		spin_unlock(&global_irq_lock);
+
+		/*
+		 * Wait for everybody else to go away and release
+		 * their things before trying to get the lock again.
+		 */
+		for (;;) {
+			STUCK;
+			if (atomic_read(&global_irq_count))
+				continue;
+			if (*((unsigned char *)&global_irq_lock))
+				continue;
+			if (spin_trylock(&global_irq_lock))
+				break;
+		}
+		atomic_add(local_count, &global_irq_count);
+	}
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 10000000
+
+#undef STUCK
+#define STUCK \
+if (!--stuck) {printk("get_irqlock stuck at %08lx, waiting for %08lx\n", where, previous_irqholder); stuck = INIT_STUCK;}
+
+static inline void get_irqlock(int cpu, unsigned long where)
+{
+	int stuck = INIT_STUCK;
+
+	if (!spin_trylock(&global_irq_lock)) {
+		/* do we already hold the lock? */
+		if ((unsigned char) cpu == global_irq_holder)
+			return;
+		/* Uhhuh.. Somebody else got it. Wait.. */
+		do {
+			do {
+				STUCK;
+				barrier();
+			} while (*((unsigned char *)&global_irq_lock));
+		} while (!spin_trylock(&global_irq_lock));
+	}
+	/*
+	 * Ok, we got the lock bit.
+	 * But that's actually just the easy part.. Now
+	 * we need to make sure that nobody else is running
+	 * in an interrupt context. 
+	 */
+	wait_on_irq(cpu, where);
+
+	/*
+	 * Finally.
+	 */
+	global_irq_holder = cpu;
+	previous_irqholder = where;
+}
+
+void __global_cli(void)
+{
+	int cpu = smp_processor_id();
+	unsigned long where;
+
+	__asm__("mov %%i7, %0" : "=r" (where));
+	__cli();
+	get_irqlock(cpu, where);
+}
+
+void __global_sti(void)
+{
+	release_irqlock(smp_processor_id());
+	__sti();
+}
+
+unsigned long __global_save_flags(void)
+{
+	return global_irq_holder == (unsigned char) smp_processor_id();
+}
+
+void __global_restore_flags(unsigned long flags)
+{
+	if(flags & 1) {
+		__global_cli();
+	} else {
+		/* release_irqlock() */
+		if(global_irq_holder == smp_processor_id()) {
+			global_irq_holder = NO_PROC_ID;
+			spin_unlock(&global_irq_lock);
+		}
+		if(!(flags & 2))
+			__sti();
+	}
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 200000000
+
+#undef STUCK
+#define STUCK \
+if (!--stuck) {printk("irq_enter stuck (irq=%d, cpu=%d, global=%d)\n",irq,cpu,global_irq_holder); stuck = INIT_STUCK;}
+
+#define VERBOSE_IRQLOCK_DEBUGGING
+
+void irq_enter(int cpu, int irq, void *_opaque)
+{
+#ifdef VERBOSE_IRQLOCK_DEBUGGING
+	extern void smp_show_backtrace_all_cpus(void);
+#endif
+	int stuck = INIT_STUCK;
+
+	hardirq_enter(cpu);
+	barrier();
+	while (*((unsigned char *)&global_irq_lock)) {
+		if ((unsigned char) cpu == global_irq_holder) {
+			struct pt_regs *regs = _opaque;
+			int sbh_cnt = atomic_read(&__sparc_bh_counter);
+			int globl_locked = *((unsigned char *)&global_irq_lock);
+			int globl_icount = atomic_read(&global_irq_count);
+			int local_count = local_irq_count[cpu];
+			unsigned long pc = regs->pc;
+
+			/* It is very important that we load the state variables
+			 * before we do the first call to printk() as printk()
+			 * could end up changing them...
+			 */
+
+			printk("CPU[%d]: BAD! Local IRQ's enabled, global disabled "
+			       "interrupt at PC[%08lx]\n", cpu, pc);
+			printk("CPU[%d]: bhcnt[%d] glocked[%d] gicnt[%d] licnt[%d]\n",
+			       cpu, sbh_cnt, globl_locked, globl_icount, local_count);
+#ifdef VERBOSE_IRQLOCK_DEBUGGING
+			printk("Performing backtrace on all cpus, write this down!\n");
+			smp_show_backtrace_all_cpus();
+#endif
+			break;
+		}
+		STUCK;
+		barrier();
+	}
+}
+
+void irq_exit(int cpu, int irq)
+{
+	hardirq_exit(cpu);
+	release_irqlock(cpu);
+}
+
+#endif /* DEBUG_IRQLOCK */
+
 /* There has to be a better way. */
-/* XXX Must write faster version in irqlock.S -DaveM */
 void synchronize_irq(void)
 {
 	int cpu = smp_processor_id();
@@ -371,7 +543,7 @@ void handler_irq(int irq, struct pt_regs * regs)
 	if(irq < 10)
 		smp_irq_rotate(cpu);
 #endif
-	irq_enter(cpu, cpu_irq);
+	irq_enter(cpu, cpu_irq, regs);
 	action = *(cpu_irq + irq_action);
 	kstat.interrupts[cpu_irq]++;
 	do {
@@ -392,7 +564,7 @@ void sparc_floppy_irq(int irq, void *dev_id, struct pt_regs *regs)
 	int cpu = smp_processor_id();
 
 	disable_pil_irq(irq);
-	irq_enter(cpu, irq);
+	irq_enter(cpu, irq, regs);
 	floppy_interrupt(irq, dev_id, regs);
 	irq_exit(cpu, irq);
 	enable_pil_irq(irq);
diff --git a/arch/sparc/kernel/process.c b/arch/sparc/kernel/process.c
index 96c57e963..1adc9e817 100644
--- a/arch/sparc/kernel/process.c
+++ b/arch/sparc/kernel/process.c
@@ -1,4 +1,4 @@
-/*  $Id: process.c,v 1.93 1997/04/11 08:55:40 davem Exp $
+/*  $Id: process.c,v 1.98 1997/05/14 20:44:54 davem Exp $
  *  linux/arch/sparc/kernel/process.c
  *
  *  Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -41,6 +41,8 @@
 
 extern void fpsave(unsigned long *, unsigned long *, void *, unsigned long *);
 
+struct task_struct *current_set[NR_CPUS] = {&init_task, };
+
 #ifndef __SMP__
 
 #define SUN4C_FAULT_HIGH 100
@@ -192,6 +194,37 @@ void show_regwindow(struct reg_window *rw)
 	       rw->ins[4], rw->ins[5], rw->ins[6], rw->ins[7]);
 }
 
+static spinlock_t sparc_backtrace_lock = SPIN_LOCK_UNLOCKED;
+
+void show_backtrace(void)
+{
+	struct reg_window *rw;
+	unsigned long flags;
+	unsigned long fp;
+	int cpu = smp_processor_id();
+
+	spin_lock_irqsave(&sparc_backtrace_lock, flags);
+	__asm__ __volatile__("mov %%i6, %0" : "=r" (fp));
+	rw = (struct reg_window *) fp;
+	while(rw) {
+		printk("CPU[%d]: ARGS[%08lx,%08lx,%08lx,%08lx,%08lx,%08lx] "
+		       "FP[%08lx] CALLER[%08lx]\n", cpu,
+		       rw->ins[0], rw->ins[1], rw->ins[2], rw->ins[3],
+		       rw->ins[4], rw->ins[5],
+		       rw->ins[6],
+		       rw->ins[7]);
+		rw = (struct reg_window *) rw->ins[6];
+	}
+	spin_unlock_irqrestore(&sparc_backtrace_lock, flags);
+}
+
+#ifdef __SMP__
+void smp_show_backtrace_all_cpus(void)
+{
+	xc0((smpfunc_t) show_backtrace);
+}
+#endif
+
 void show_stackframe(struct sparc_stackf *sf)
 {
 	unsigned long size;
@@ -441,12 +474,12 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 
 	if(regs->psr & PSR_PS)
 		stack_offset -= REGWIN_SZ;
-	childregs = ((struct pt_regs *) (p->kernel_stack_page + stack_offset));
+	childregs = ((struct pt_regs *) (((unsigned long)p) + stack_offset));
 	copy_regs(childregs, regs);
 	new_stack = (((struct reg_window *) childregs) - 1);
 	copy_regwin(new_stack, (((struct reg_window *) regs) - 1));
 
-	p->tss.ksp = p->saved_kernel_stack = (unsigned long) new_stack;
+	p->tss.ksp = (unsigned long) new_stack;
 #ifdef __SMP__
 	p->tss.kpc = (((unsigned long) ret_from_smpfork) - 0x8);
 	p->tss.kpsr = current->tss.fork_kpsr | PSR_PIL;
@@ -467,7 +500,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 		p->tss.flags &= ~SPARC_FLAG_KTHREAD;
 		p->tss.current_ds = USER_DS;
 
-		if (sp != current->tss.kregs->u_regs[UREG_FP]) {
+		if (sp != regs->u_regs[UREG_FP]) {
 			struct sparc_stackf *childstack;
 			struct sparc_stackf *parentstack;
 
@@ -475,9 +508,8 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 			 * This is a clone() call with supplied user stack.
 			 * Set some valid stack frames to give to the child.
 			 */
-			childstack = (struct sparc_stackf *)sp;
-			parentstack = (struct sparc_stackf *)
-					current->tss.kregs->u_regs[UREG_FP];
+			childstack = (struct sparc_stackf *) sp;
+			parentstack = (struct sparc_stackf *) regs->u_regs[UREG_FP];
 
 #if 0
 			printk("clone: parent stack:\n");
diff --git a/arch/sparc/kernel/ptrace.c b/arch/sparc/kernel/ptrace.c
index 647081eda..732b3006d 100644
--- a/arch/sparc/kernel/ptrace.c
+++ b/arch/sparc/kernel/ptrace.c
@@ -24,18 +24,6 @@
 
 #define MAGIC_CONSTANT 0x80000000
 
-/* change a pid into a task struct. */
-static inline struct task_struct * get_task(int pid)
-{
-	int i;
-
-	for (i = 1; i < NR_TASKS; i++) {
-		if (task[i] != NULL && (task[i]->pid == pid))
-			return task[i];
-	}
-	return NULL;
-}
-
 /*
  * This routine gets a long from any process space by following the page
  * tables. NOTE! You should check that the long isn't on a page boundary,
@@ -53,7 +41,7 @@ static unsigned long get_long(struct task_struct * tsk,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (pgd_none(*pgdir)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -63,7 +51,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -73,7 +61,7 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 0);
+		handle_mm_fault(tsk, vma, addr, 0);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
@@ -106,7 +94,7 @@ static void put_long(struct task_struct * tsk, struct vm_area_struct * vma,
 repeat:
 	pgdir = pgd_offset(vma->vm_mm, addr);
 	if (!pgd_present(*pgdir)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pgd_bad(*pgdir)) {
@@ -116,7 +104,7 @@ repeat:
 	}
 	pgmiddle = pmd_offset(pgdir, addr);
 	if (pmd_none(*pgmiddle)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	if (pmd_bad(*pgmiddle)) {
@@ -126,12 +114,12 @@ repeat:
 	}
 	pgtable = pte_offset(pgmiddle, addr);
 	if (!pte_present(*pgtable)) {
-		do_no_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 	page = pte_page(*pgtable);
 	if (!pte_write(*pgtable)) {
-		do_wp_page(tsk, vma, addr, 1);
+		handle_mm_fault(tsk, vma, addr, 1);
 		goto repeat;
 	}
 /* this is a hack for non-kernel-mapped video buffers and similar */
@@ -533,7 +521,7 @@ asmlinkage void do_ptrace(struct pt_regs *regs)
 		goto out;
 	}
 #endif
-	if(!(child = get_task(pid))) {
+	if(!(child = find_task_by_pid(pid))) {
 		pt_error_return(regs, ESRCH);
 		goto out;
 	}
diff --git a/arch/sparc/kernel/setup.c b/arch/sparc/kernel/setup.c
index f2af0b763..aa204db06 100644
--- a/arch/sparc/kernel/setup.c
+++ b/arch/sparc/kernel/setup.c
@@ -1,4 +1,4 @@
-/*  $Id: setup.c,v 1.83 1997/04/01 02:21:49 davem Exp $
+/*  $Id: setup.c,v 1.84 1997/05/08 17:45:16 davem Exp $
  *  linux/arch/sparc/kernel/setup.c
  *
  *  Copyright (C) 1995  David S. Miller (davem@caip.rutgers.edu)
@@ -83,7 +83,8 @@ void prom_sync_me(void)
 
 #ifdef __SMP__
 	global_irq_holder = NO_PROC_ID;
-	global_irq_lock = global_bh_lock = 0;
+	*((unsigned char *)&global_irq_lock) = 0;
+	*((unsigned char *)&global_bh_lock) = 0;
 #endif
 	__save_and_cli(flags);
 	__asm__ __volatile__("rd %%tbr, %0\n\t" : "=r" (prom_tbr));
diff --git a/arch/sparc/kernel/signal.c b/arch/sparc/kernel/signal.c
index 8a07ae0be..fd5fa634e 100644
--- a/arch/sparc/kernel/signal.c
+++ b/arch/sparc/kernel/signal.c
@@ -1,4 +1,4 @@
-/*  $Id: signal.c,v 1.73 1997/04/16 05:56:05 davem Exp $
+/*  $Id: signal.c,v 1.74 1997/05/15 19:57:09 davem Exp $
  *  linux/arch/sparc/kernel/signal.c
  *
  *  Copyright (C) 1991, 1992  Linus Torvalds
@@ -780,8 +780,10 @@ asmlinkage int do_signal(unsigned long oldmask, struct pt_regs * regs,
 			case SIGQUIT: case SIGILL: case SIGTRAP:
 			case SIGABRT: case SIGFPE: case SIGSEGV: case SIGBUS:
 				if(current->binfmt && current->binfmt->core_dump) {
+					lock_kernel();
 					if(current->binfmt->core_dump(signr, regs))
 						signr |= 0x80;
+					unlock_kernel();
 				}
 #ifdef DEBUG_SIGNALS
 				/* Very useful to debug dynamic linker problems */
diff --git a/arch/sparc/kernel/smp.c b/arch/sparc/kernel/smp.c
index 6622f88f0..beef2df14 100644
--- a/arch/sparc/kernel/smp.c
+++ b/arch/sparc/kernel/smp.c
@@ -4,16 +4,18 @@
  */
 
 #include <asm/head.h>
-#include <asm/ptrace.h>
-#include <asm/atomic.h>
 
 #include <linux/kernel.h>
+#include <linux/sched.h>
 #include <linux/tasks.h>
 #include <linux/smp.h>
 #include <linux/smp_lock.h>
 #include <linux/interrupt.h>
 #include <linux/kernel_stat.h>
 
+#include <asm/ptrace.h>
+#include <asm/atomic.h>
+
 #include <asm/delay.h>
 #include <asm/irq.h>
 #include <asm/page.h>
@@ -24,6 +26,9 @@
 #include <asm/hardirq.h>
 #include <asm/softirq.h>
 
+#define __KERNEL_SYSCALLS__
+#include <linux/unistd.h>
+
 #define IRQ_RESCHEDULE		13
 #define IRQ_STOP_CPU		14
 #define IRQ_CROSS_CALL		15
@@ -33,6 +38,9 @@ extern int linux_num_cpus;
 
 extern void calibrate_delay(void);
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
+
 volatile int smp_processors_ready = 0;
 
 unsigned long cpu_present_map = 0;
@@ -118,16 +126,6 @@ void smp_store_cpu_info(int id)
 	cpu_data[id].udelay_val = loops_per_sec; /* this is it on sparc. */
 }
 
-/*
- *	Architecture specific routine called by the kernel just before init is
- *	fired off. This allows the BP to have everything in order [we hope].
- *	At the end of this all the AP's will hit the system scheduling and off
- *	we go. Each AP will load the system gdt's and jump through the kernel
- *	init into idle(). At this point the scheduler will one day take over 
- * 	and give them jobs to do. smp_callin is a standard routine
- *	we use to track CPU's as they power up.
- */
-
 void smp_commence(void)
 {
 	/*
@@ -144,7 +142,7 @@ static void smp_setup_percpu_timer(void);
 
 void smp_callin(void)
 {
-	int cpuid = smp_processor_id();
+	int cpuid = hard_smp_processor_id();
 
 	local_flush_cache_all();
 	local_flush_tlb_all();
@@ -183,6 +181,25 @@ void smp_callin(void)
 	__sti();
 }
 
+extern int cpu_idle(void *unused);
+extern void init_IRQ(void);
+
+/* Only broken Intel needs this, thus it should not even be referenced
+ * globally...
+ */
+void initialize_secondary(void)
+{
+}
+
+/* Activate a secondary processor. */
+int start_secondary(void *unused)
+{
+	trap_init();
+	init_IRQ();
+	smp_callin();
+	return cpu_idle(NULL);
+}
+
 void cpu_panic(void)
 {
 	printk("CPU[%d]: Returns from cpu_idle!\n", smp_processor_id());
@@ -220,6 +237,7 @@ void smp_boot_cpus(void)
 	cpu_number_map[boot_cpu_id] = 0;
 	cpu_logical_map[0] = boot_cpu_id;
 	klock_info.akp = boot_cpu_id;
+	current->processor = boot_cpu_id;
 	smp_store_cpu_info(boot_cpu_id);
 	set_irq_udt(mid_xlate[boot_cpu_id]);
 	smp_setup_percpu_timer();
@@ -233,10 +251,19 @@ void smp_boot_cpus(void)
 		if(cpu_present_map & (1 << i)) {
 			extern unsigned long sparc_cpu_startup;
 			unsigned long *entry = &sparc_cpu_startup;
+			struct task_struct *p;
 			int timeout;
 
+			/* Cook up an idler for this guy. */
+			kernel_thread(start_secondary, NULL, CLONE_PID);
+
+			p = task[++cpucount];
+
+			p->processor = i;
+			current_set[i] = p;
+
 			/* See trampoline.S for details... */
-			entry += ((i-1) * 6);
+			entry += ((i-1) * 3);
 
 			/* whirrr, whirrr, whirrrrrrrrr... */
 			printk("Starting CPU %d at %p\n", i, entry);
@@ -253,10 +280,10 @@ void smp_boot_cpus(void)
 			}
 			if(cpu_callin_map[i]) {
 				/* Another "Red Snapper". */
-				cpucount++;
 				cpu_number_map[i] = i;
 				cpu_logical_map[i] = i;
 			} else {
+				cpucount--;
 				printk("Processor %d is stuck.\n", i);
 			}
 		}
@@ -364,26 +391,17 @@ struct smp_funcall {
 	unsigned long processors_out[NR_CPUS]; /* Set when ipi exited. */
 } ccall_info;
 
-/* Returns failure code if for example any of the cpu's failed to respond
- * within a certain timeout period.
- */
-
-#define CCALL_TIMEOUT   5000000 /* enough for initial testing */
-
 static spinlock_t cross_call_lock = SPIN_LOCK_UNLOCKED;
 
 /* Cross calls must be serialized, at least currently. */
 void smp_cross_call(smpfunc_t func, unsigned long arg1, unsigned long arg2,
 		    unsigned long arg3, unsigned long arg4, unsigned long arg5)
 {
-	unsigned long me = smp_processor_id();
-	unsigned long flags, mask;
-	int i, timeout;
-
 	if(smp_processors_ready) {
-		__save_flags(flags);
-		__cli();
-		spin_lock(&cross_call_lock);
+		register int ncpus = smp_num_cpus;
+		unsigned long flags;
+
+		spin_lock_irqsave(&cross_call_lock, flags);
 
 		/* Init function glue. */
 		ccall_info.func = func;
@@ -393,56 +411,47 @@ void smp_cross_call(smpfunc_t func, unsigned long arg1, unsigned long arg2,
 		ccall_info.arg4 = arg4;
 		ccall_info.arg5 = arg5;
 
-		/* Init receive/complete mapping. */
-		for(i = 0; i < smp_num_cpus; i++) {
-			ccall_info.processors_in[i] = 0;
-			ccall_info.processors_out[i] = 0;
+		/* Init receive/complete mapping, plus fire the IPI's off. */
+		{
+			register void (*send_ipi)(int,int) = set_cpu_int;
+			register unsigned long mask;
+			register int i;
+
+			mask = (cpu_present_map & ~(1 << smp_processor_id()));
+			for(i = 0; i < ncpus; i++) {
+				if(mask & (1 << i)) {
+					ccall_info.processors_in[i] = 0;
+					ccall_info.processors_out[i] = 0;
+					send_ipi(mid_xlate[i], IRQ_CROSS_CALL);
+				} else {
+					ccall_info.processors_in[i] = 1;
+					ccall_info.processors_out[i] = 1;
+				}
+			}
 		}
-		ccall_info.processors_in[me] = 1;
-		ccall_info.processors_out[me] = 1;
 
-		/* Fire it off. */
-		mask = (cpu_present_map & ~(1 << me));
-		for(i = 0; i < 4; i++) {
-			if(mask & (1 << i))
-				set_cpu_int(mid_xlate[i], IRQ_CROSS_CALL);
-		}
+		/* First, run local copy. */
+		func(arg1, arg2, arg3, arg4, arg5);
 
-		/* For debugging purposes right now we can timeout
-		 * on both callin and callexit.
-		 */
-		timeout = CCALL_TIMEOUT;
-		for(i = 0; i < smp_num_cpus; i++) {
-			while(!ccall_info.processors_in[i] && timeout-- > 0)
-				barrier();
-			if(!ccall_info.processors_in[i])
-				goto procs_time_out;
-		}
+		{
+			register int i;
 
-		/* Run local copy. */
-		func(arg1, arg2, arg3, arg4, arg5);
+			i = 0;
+			do {
+				while(!ccall_info.processors_in[i])
+					barrier();
+			} while(++i < ncpus);
 
-		/* Spin on proc dispersion. */
-		timeout = CCALL_TIMEOUT;
-		for(i = 0; i < smp_num_cpus; i++) {
-			while(!ccall_info.processors_out[i] && timeout-- > 0)
-				barrier();
-			if(!ccall_info.processors_out[i])
-				goto procs_time_out;
+			i = 0;
+			do {
+				while(!ccall_info.processors_out[i])
+					barrier();
+			} while(++i < ncpus);
 		}
-		spin_unlock(&cross_call_lock);
-		__restore_flags(flags);
-		return; /* made it... */
-
-procs_time_out:
-		printk("smp: Wheee, penguin drops off the bus\n");
-		spin_unlock(&cross_call_lock);
-		__restore_flags(flags);
-		return; /* why me... why me... */
-	}
 
-	/* Just need to run local copy. */
-	func(arg1, arg2, arg3, arg4, arg5);
+		spin_unlock_irqrestore(&cross_call_lock, flags);
+	} else
+		func(arg1, arg2, arg3, arg4, arg5); /* Just need to run local copy. */
 }
 
 void smp_flush_cache_all(void)
@@ -453,43 +462,98 @@ void smp_flush_tlb_all(void)
 
 void smp_flush_cache_mm(struct mm_struct *mm)
 { 
-	if(mm->context != NO_CONTEXT)
-		xc1((smpfunc_t) local_flush_cache_mm, (unsigned long) mm);
+	if(mm->context != NO_CONTEXT) {
+		if(mm->cpu_vm_mask == (1 << smp_processor_id()))
+			local_flush_cache_mm(mm);
+		else
+			xc1((smpfunc_t) local_flush_cache_mm, (unsigned long) mm);
+	}
 }
 
 void smp_flush_tlb_mm(struct mm_struct *mm)
 {
-	if(mm->context != NO_CONTEXT)
-		xc1((smpfunc_t) local_flush_tlb_mm, (unsigned long) mm);
+	if(mm->context != NO_CONTEXT) {
+		if(mm->cpu_vm_mask == (1 << smp_processor_id())) {
+			local_flush_tlb_mm(mm);
+		} else {
+			xc1((smpfunc_t) local_flush_tlb_mm, (unsigned long) mm);
+			if(mm->count == 1 && current->mm == mm)
+				mm->cpu_vm_mask = (1 << smp_processor_id());
+		}
+	}
 }
 
 void smp_flush_cache_range(struct mm_struct *mm, unsigned long start,
 			   unsigned long end)
 {
-	if(mm->context != NO_CONTEXT)
-		xc3((smpfunc_t) local_flush_cache_range, (unsigned long) mm,
-		    start, end);
+	if(mm->context != NO_CONTEXT) {
+		if(mm->cpu_vm_mask == (1 << smp_processor_id()))
+			local_flush_cache_range(mm, start, end);
+		else
+			xc3((smpfunc_t) local_flush_cache_range, (unsigned long) mm,
+			    start, end);
+	}
 }
 
 void smp_flush_tlb_range(struct mm_struct *mm, unsigned long start,
 			 unsigned long end)
 {
-	if(mm->context != NO_CONTEXT)
-		xc3((smpfunc_t) local_flush_tlb_range, (unsigned long) mm,
-		    start, end);
+	if(mm->context != NO_CONTEXT) {
+		if(mm->cpu_vm_mask == (1 << smp_processor_id()))
+			local_flush_tlb_range(mm, start, end);
+		else
+			xc3((smpfunc_t) local_flush_tlb_range, (unsigned long) mm,
+			    start, end);
+	}
 }
 
 void smp_flush_cache_page(struct vm_area_struct *vma, unsigned long page)
-{ xc2((smpfunc_t) local_flush_cache_page, (unsigned long) vma, page); }
+{
+	struct mm_struct *mm = vma->vm_mm;
+
+	if(mm->context != NO_CONTEXT) {
+		if(mm->cpu_vm_mask == (1 << smp_processor_id()))
+			local_flush_cache_page(vma, page);
+		else
+			xc2((smpfunc_t) local_flush_cache_page,
+			    (unsigned long) vma, page);
+	}
+}
 
 void smp_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
-{ xc2((smpfunc_t) local_flush_tlb_page, (unsigned long) vma, page); }
+{
+	struct mm_struct *mm = vma->vm_mm;
+
+	if(mm->context != NO_CONTEXT) {
+		if(mm->cpu_vm_mask == (1 << smp_processor_id()))
+			local_flush_tlb_page(vma, page);
+		else
+			xc2((smpfunc_t) local_flush_tlb_page, (unsigned long) vma, page);
+	}
+}
 
 void smp_flush_page_to_ram(unsigned long page)
-{ xc1((smpfunc_t) local_flush_page_to_ram, page); }
+{
+	/* Current theory is that those who call this are the one's
+	 * who have just dirtied their cache with the pages contents
+	 * in kernel space, therefore we only run this on local cpu.
+	 *
+	 * XXX This experiment failed, research further... -DaveM
+	 */
+#if 1
+	xc1((smpfunc_t) local_flush_page_to_ram, page);
+#else
+	local_flush_page_to_ram(page);
+#endif
+}
 
 void smp_flush_sig_insns(struct mm_struct *mm, unsigned long insn_addr)
-{ xc2((smpfunc_t) local_flush_sig_insns, (unsigned long) mm, insn_addr); }
+{
+	if(mm->cpu_vm_mask == (1 << smp_processor_id()))
+		local_flush_sig_insns(mm, insn_addr);
+	else
+		xc2((smpfunc_t) local_flush_sig_insns, (unsigned long) mm, insn_addr);
+}
 
 /* Reschedule call back. */
 void smp_reschedule_irq(void)
diff --git a/arch/sparc/kernel/sparc_ksyms.c b/arch/sparc/kernel/sparc_ksyms.c
index 645aafb35..38896ab22 100644
--- a/arch/sparc/kernel/sparc_ksyms.c
+++ b/arch/sparc/kernel/sparc_ksyms.c
@@ -1,4 +1,4 @@
-/* $Id: sparc_ksyms.c,v 1.56 1997/04/18 05:44:35 davem Exp $
+/* $Id: sparc_ksyms.c,v 1.59 1997/05/08 17:45:20 davem Exp $
  * arch/sparc/kernel/ksyms.c: Sparc specific ksyms support.
  *
  * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
@@ -38,6 +38,7 @@
 #include <asm/dma.h>
 #endif
 #include <asm/a.out.h>
+#include <asm/spinlock.h>
 
 struct poll {
 	int fd;
@@ -50,9 +51,9 @@ extern int svr4_setcontext (svr4_ucontext_t *, struct pt_regs *);
 extern unsigned long sunos_mmap(unsigned long, unsigned long, unsigned long,
 				unsigned long, unsigned long, unsigned long);
 void _sigpause_common (unsigned int set, struct pt_regs *);
-extern void __copy_1page(void *, const void *);
+extern void (*__copy_1page)(void *, const void *);
 extern void __memmove(void *, const void *, __kernel_size_t);
-extern void *bzero_1page(void *);
+extern void (*bzero_1page)(void *);
 extern void *__bzero(void *, size_t);
 extern void *__memscan_zero(void *, size_t);
 extern void *__memscan_generic(void *, int, size_t);
@@ -87,17 +88,47 @@ EXPORT_SYMBOL(klock_info);
 EXPORT_SYMBOL_PRIVATE(_lock_kernel);
 EXPORT_SYMBOL_PRIVATE(_unlock_kernel);
 EXPORT_SYMBOL_PRIVATE(_spinlock_waitfor);
+#ifdef SPIN_LOCK_DEBUG
+EXPORT_SYMBOL(_spin_lock);
+EXPORT_SYMBOL(_spin_unlock);
+EXPORT_SYMBOL(_spin_trylock);
+EXPORT_SYMBOL(_spin_lock_irq);
+EXPORT_SYMBOL(_spin_unlock_irq);
+EXPORT_SYMBOL(_spin_lock_irqsave);
+EXPORT_SYMBOL(_spin_unlock_irqrestore);
+EXPORT_SYMBOL(_read_lock);
+EXPORT_SYMBOL(_read_unlock);
+EXPORT_SYMBOL(_read_lock_irq);
+EXPORT_SYMBOL(_read_unlock_irq);
+EXPORT_SYMBOL(_read_lock_irqsave);
+EXPORT_SYMBOL(_read_unlock_irqrestore);
+EXPORT_SYMBOL(_write_lock);
+EXPORT_SYMBOL(_write_unlock);
+EXPORT_SYMBOL(_write_lock_irq);
+EXPORT_SYMBOL(_write_unlock_irq);
+EXPORT_SYMBOL(_write_lock_irqsave);
+EXPORT_SYMBOL(_write_unlock_irqrestore);
+#else
 EXPORT_SYMBOL_PRIVATE(_rw_read_enter);
 EXPORT_SYMBOL_PRIVATE(_rw_read_exit);
 EXPORT_SYMBOL_PRIVATE(_rw_write_enter);
+#endif
 EXPORT_SYMBOL(__sparc_bh_counter);
 #ifdef __SMP__
+#ifdef DEBUG_IRQLOCK
+EXPORT_SYMBOL(irq_enter);
+EXPORT_SYMBOL(irq_exit);
+EXPORT_SYMBOL(__global_restore_flags);
+EXPORT_SYMBOL(__global_sti);
+EXPORT_SYMBOL(__global_cli);
+#else
 EXPORT_SYMBOL_PRIVATE(_irq_enter);
 EXPORT_SYMBOL_PRIVATE(_irq_exit);
 EXPORT_SYMBOL_PRIVATE(_global_restore_flags);
 EXPORT_SYMBOL_PRIVATE(_global_sti);
 EXPORT_SYMBOL_PRIVATE(_global_cli);
 #endif
+#endif
 
 EXPORT_SYMBOL(page_offset);
 EXPORT_SYMBOL(stack_top);
diff --git a/arch/sparc/kernel/sys_sunos.c b/arch/sparc/kernel/sys_sunos.c
index 5f5b19504..3fafe7f9c 100644
--- a/arch/sparc/kernel/sys_sunos.c
+++ b/arch/sparc/kernel/sys_sunos.c
@@ -1,4 +1,4 @@
-/* $Id: sys_sunos.c,v 1.78 1997/04/16 05:56:12 davem Exp $
+/* $Id: sys_sunos.c,v 1.79 1997/04/23 23:01:15 ecd Exp $
  * sys_sunos.c: SunOS specific syscall compatibility support.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -20,6 +20,7 @@
 #include <linux/resource.h>
 #include <linux/ipc.h>
 #include <linux/shm.h>
+#include <linux/msg.h>
 #include <linux/sem.h>
 #include <linux/signal.h>
 #include <linux/uio.h>
@@ -1031,7 +1032,48 @@ asmlinkage int sunos_semsys(int op, unsigned long arg1, unsigned long arg2,
 	unlock_kernel();
 	return ret;
 }
-    
+
+extern asmlinkage int sys_msgget (key_t key, int msgflg);
+extern asmlinkage int sys_msgrcv (int msqid, struct msgbuf *msgp,
+				  size_t msgsz, long msgtyp, int msgflg);
+extern asmlinkage int sys_msgsnd (int msqid, struct msgbuf *msgp,
+				  size_t msgsz, int msgflg);
+extern asmlinkage int sys_msgctl (int msqid, int cmd, struct msqid_ds *buf);
+
+asmlinkage int sunos_msgsys(int op, unsigned long arg1, unsigned long arg2,
+			    unsigned long arg3, unsigned long arg4)
+{
+	struct sparc_stackf *sp;
+	unsigned long arg5;
+	int rval;
+
+	lock_kernel();
+	switch(op) {
+	case 0:
+		rval = sys_msgget((key_t)arg1, (int)arg2);
+		break;
+	case 1:
+		rval = sys_msgctl((int)arg1, (int)arg2,
+				  (struct msqid_ds *)arg3);
+		break;
+	case 2:
+		sp = (struct sparc_stackf *)current->tss.kregs->u_regs[UREG_FP];
+		arg5 = sp->xxargs[0];
+		rval = sys_msgrcv((int)arg1, (struct msgbuf *)arg2,
+				  (size_t)arg3, (long)arg4, (int)arg5);
+		break;
+	case 3:
+		rval = sys_msgsnd((int)arg1, (struct msgbuf *)arg2,
+				  (size_t)arg3, (int)arg4);
+		break;
+	default:
+		rval = -EINVAL;
+		break;
+	}
+	unlock_kernel();
+	return rval;
+}
+
 extern asmlinkage int sys_shmat (int shmid, char *shmaddr, int shmflg, ulong *raddr);
 extern asmlinkage int sys_shmctl (int shmid, int cmd, struct shmid_ds *buf);
 extern asmlinkage int sys_shmdt (char *shmaddr);
diff --git a/arch/sparc/kernel/systbls.S b/arch/sparc/kernel/systbls.S
index 61b41056b..a166e3490 100644
--- a/arch/sparc/kernel/systbls.S
+++ b/arch/sparc/kernel/systbls.S
@@ -1,4 +1,4 @@
-/* $Id: systbls.S,v 1.60 1997/04/19 08:52:15 jj Exp $
+/* $Id: systbls.S,v 1.62 1997/04/23 23:01:08 ecd Exp $
  * systbls.S: System call entry point tables for OS compatibility.
  *            The native Linux system call table lives here also.
  *
@@ -106,15 +106,26 @@ C_LABEL(sys_call_table):
 	.long C_LABEL(sys_nis_syscall), C_LABEL(sys_stime), C_LABEL(sys_nis_syscall)
 	.long C_LABEL(sys_nis_syscall), C_LABEL(sys_llseek)
 	/* "We are the Knights of the Forest of Ni!!" */
-	.long C_LABEL(sys_mlock), C_LABEL(sys_munlock), C_LABEL(sys_mlockall)
-	.long C_LABEL(sys_munlockall), C_LABEL(sys_sched_setparam)
-	.long C_LABEL(sys_sched_getparam), C_LABEL(sys_nis_syscall), C_LABEL(sys_nis_syscall)
-	.long C_LABEL(sys_nis_syscall), C_LABEL(sys_sched_get_priority_max), C_LABEL(sys_sched_get_priority_min)
-	.long C_LABEL(sys_sched_rr_get_interval), C_LABEL(sys_nanosleep)
+	.long C_LABEL(sys_mlock)
+	.long C_LABEL(sys_munlock)
+	.long C_LABEL(sys_mlockall)
+/*240*/	.long C_LABEL(sys_munlockall)
+	.long C_LABEL(sys_sched_setparam)
+	.long C_LABEL(sys_sched_getparam)
+	.long C_LABEL(sys_sched_setscheduler)
+	.long C_LABEL(sys_sched_getscheduler)
+/*245*/	.long C_LABEL(sys_sched_yield)
+	.long C_LABEL(sys_sched_get_priority_max)
+	.long C_LABEL(sys_sched_get_priority_min)
+	.long C_LABEL(sys_sched_rr_get_interval)
+	.long C_LABEL(sys_nanosleep)
 /*250*/	.long C_LABEL(sys_mremap)
 	.long C_LABEL(sys_sysctl)
-	.long C_LABEL(sys_getsid), C_LABEL(sys_fdatasync), C_LABEL(sys_nfsservctl)
-	.long C_LABEL(sys_aplib), C_LABEL(sys_nis_syscall)
+	.long C_LABEL(sys_getsid)
+	.long C_LABEL(sys_fdatasync)
+	.long C_LABEL(sys_nfsservctl)
+/*255*/	.long C_LABEL(sys_aplib)
+	.long C_LABEL(sys_nis_syscall)
 
 	/* Now the SunOS syscall table. */
 
@@ -179,7 +190,7 @@ C_LABEL(sunos_sys_table):
 	.long C_LABEL(sunos_getdomainname), C_LABEL(sys_setdomainname)
 	.long C_LABEL(sunos_nosys), C_LABEL(sys_quotactl), C_LABEL(sunos_nosys)
 	.long C_LABEL(sunos_mount), C_LABEL(sys_ustat), C_LABEL(sunos_semsys)
-	.long C_LABEL(sunos_nosys), C_LABEL(sunos_shmsys), C_LABEL(sunos_audit)
+	.long C_LABEL(sunos_msgsys), C_LABEL(sunos_shmsys), C_LABEL(sunos_audit)
 	.long C_LABEL(sunos_nosys), C_LABEL(sunos_getdents), C_LABEL(sys_setsid)
 	.long C_LABEL(sys_fchdir), C_LABEL(sunos_nosys), C_LABEL(sunos_nosys)
 	.long C_LABEL(sunos_nosys), C_LABEL(sunos_nosys), C_LABEL(sunos_nosys)
diff --git a/arch/sparc/kernel/tadpole.c b/arch/sparc/kernel/tadpole.c
index bea6336c5..03fe3cff9 100644
--- a/arch/sparc/kernel/tadpole.c
+++ b/arch/sparc/kernel/tadpole.c
@@ -4,6 +4,8 @@
  */
 
 #include <linux/string.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
 
 #include <asm/asi.h>
 #include <asm/oplib.h>
diff --git a/arch/sparc/kernel/trampoline.S b/arch/sparc/kernel/trampoline.S
index c59a2531a..9ee5bd14a 100644
--- a/arch/sparc/kernel/trampoline.S
+++ b/arch/sparc/kernel/trampoline.S
@@ -1,4 +1,4 @@
-/* $Id: trampoline.S,v 1.6 1997/04/14 05:38:33 davem Exp $
+/* $Id: trampoline.S,v 1.9 1997/05/01 08:53:34 davem Exp $
  * trampoline.S: SMP cpu boot-up trampoline code.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -7,6 +7,7 @@
 #include <asm/cprefix.h>
 #include <asm/head.h>
 #include <asm/psr.h>
+#include <asm/page.h>
 #include <asm/asi.h>
 #include <asm/ptrace.h>
 #include <asm/vaddrs.h>
@@ -25,27 +26,18 @@
 C_LABEL(sparc_cpu_startup):
 cpu1_startup:
 	sethi	%hi(C_LABEL(trapbase_cpu1)), %g3
-	or	%g3, %lo(C_LABEL(trapbase_cpu1)), %g3
-	sethi	%hi(C_LABEL(cpu1_stack)), %g2
-	or	%g2, %lo(C_LABEL(cpu1_stack)), %g2
 	b	1f
-	 nop	
+	 or	%g3, %lo(C_LABEL(trapbase_cpu1)), %g3
 
 cpu2_startup:
 	sethi	%hi(C_LABEL(trapbase_cpu2)), %g3
-	or	%g3, %lo(C_LABEL(trapbase_cpu2)), %g3
-	sethi	%hi(C_LABEL(cpu2_stack)), %g2
-	or	%g2, %lo(C_LABEL(cpu2_stack)), %g2
 	b	1f
-	 nop	
+	 or	%g3, %lo(C_LABEL(trapbase_cpu2)), %g3
 
 cpu3_startup:
 	sethi	%hi(C_LABEL(trapbase_cpu3)), %g3
-	or	%g3, %lo(C_LABEL(trapbase_cpu3)), %g3
-	sethi	%hi(C_LABEL(cpu3_stack)), %g2
-	or	%g2, %lo(C_LABEL(cpu3_stack)), %g2
 	b	1f
-	 nop	
+	 or	%g3, %lo(C_LABEL(trapbase_cpu3)), %g3
 
 1:
 	/* Set up a sane %psr -- PIL<0xf> S<0x1> PS<0x1> CWP<0x0> */
@@ -62,14 +54,16 @@ cpu3_startup:
 	wr	%g3, 0x0, %tbr
 	WRITE_PAUSE
 
-	/* Give ourselves a stack. */
-	set	0x2000, %g5
-	add	%g2, %g5, %g2		! end of stack
-	sub	%g2, REGWIN_SZ, %sp
-	mov	0, %fp
+	/* Give ourselves a stack and curptr. */
+	set	C_LABEL(current_set), %g5
+	srl	%g3, 10, %g4
+	and	%g4, 0xc, %g4
+	ld	[%g5 + %g4], %g6
 
-	/* Set up curptr. */
-	set	C_LABEL(init_task), %g6
+	mov	1, %sp
+	sll	%sp, (PAGE_SHIFT + 1), %sp
+	sub	%sp, REGWIN_SZ, %sp
+	add	%g6, %sp, %sp
 
 	/* Turn on traps (PSR_ET). */
 	rd	%psr, %g1
diff --git a/arch/sparc/kernel/wof.S b/arch/sparc/kernel/wof.S
index bad0088f4..890676bfb 100644
--- a/arch/sparc/kernel/wof.S
+++ b/arch/sparc/kernel/wof.S
@@ -1,4 +1,4 @@
-/* $Id: wof.S,v 1.33 1997/03/04 16:26:35 jj Exp $
+/* $Id: wof.S,v 1.36 1997/05/01 08:53:35 davem Exp $
  * wof.S: Sparc window overflow handler.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -234,15 +234,20 @@ spwin_user_stack_is_bolixed:
 spnwin_patch3:	and	%twin_tmp, 0xff, %twin_tmp	! patched on 7win Sparcs
 		st	%twin_tmp, [%curptr + AOFF_task_tss + AOFF_thread_uwinmask]
 
-	/* Jump onto kernel stack for this process... */
-	ld	[%curptr + AOFF_task_saved_kernel_stack], %sp
+	mov	1, %sp
+	sll	%sp, (PAGE_SHIFT + 1), %sp
+	sub	%sp, (TRACEREG_SZ + REGWIN_SZ), %sp
+	add	%curptr, %sp, %sp
 
 	/* Restore the saved globals and build a pt_regs frame. */
 	mov	%saved_g5, %g5
-	mov	%g6, %l4
 	mov	%saved_g6, %g6
 	STORE_PT_ALL(sp, t_psr, t_pc, t_npc, g1)
-	mov	%l4, %g6
+
+	mov	1, %g6
+	sll	%g6, (PAGE_SHIFT + 1), %g6
+	sub	%g6, (TRACEREG_SZ + REGWIN_SZ), %g6
+	sub	%sp, %g6, %g6
 
 	/* Turn on traps and call c-code to deal with it. */
 	wr	%t_psr, PSR_ET, %psr
diff --git a/arch/sparc/kernel/wuf.S b/arch/sparc/kernel/wuf.S
index 7550dd627..cb407aa69 100644
--- a/arch/sparc/kernel/wuf.S
+++ b/arch/sparc/kernel/wuf.S
@@ -1,4 +1,4 @@
-/* $Id: wuf.S,v 1.31 1997/03/04 16:26:37 jj Exp $
+/* $Id: wuf.S,v 1.34 1997/05/01 08:53:36 davem Exp $
  * wuf.S: Window underflow trap handler for the Sparc.
  *
  * Copyright (C) 1995 David S. Miller
@@ -145,14 +145,17 @@ fwin_user_stack_is_bolixed:
 	 * to the trap window and call c-code to deal with this.
 	 */
 	LOAD_CURRENT(l4, l5)
-	ld	[%l4 + AOFF_task_saved_kernel_stack], %l5
+
+	mov	1, %l5
+	sll	%l5, (PAGE_SHIFT + 1), %l5
+	sub	%l5, (TRACEREG_SZ + REGWIN_SZ), %l5
+	add	%l4, %l5, %l5
 
 	/* Store globals into pt_regs frame. */
 	STORE_PT_GLOBALS(l5)
 	STORE_PT_YREG(l5, g3)
 
-	/* Save kernel %sp in global while we change windows. */
-	mov	%l5, %g2
+	/* Save current in a global while we change windows. */
 	mov	%l4, %curptr
 
 	save	%g0, %g0, %g0
@@ -166,7 +169,10 @@ fwin_user_stack_is_bolixed:
 
 	/* LOCATION: Window 'T' */
 
-	mov	%g2, %sp		/* Jump onto kernel %sp being held */
+	mov	1, %sp
+	sll	%sp, (PAGE_SHIFT + 1), %sp
+	sub	%sp, (TRACEREG_SZ + REGWIN_SZ), %sp
+	add	%curptr, %sp, %sp
 
 	/* Build rest of pt_regs. */
 	STORE_PT_INS(sp)
diff --git a/arch/sparc/lib/Makefile b/arch/sparc/lib/Makefile
index 1bf49bf13..cefe7a851 100644
--- a/arch/sparc/lib/Makefile
+++ b/arch/sparc/lib/Makefile
@@ -1,13 +1,11 @@
-# $Id: Makefile,v 1.23 1997/04/18 05:44:39 davem Exp $
+# $Id: Makefile,v 1.24 1997/05/08 17:45:26 davem Exp $
 # Makefile for Sparc library files..
 #
 
-CFLAGS := $(CFLAGS) -ansi
-
 OBJS  = mul.o rem.o sdiv.o udiv.o umul.o urem.o ashrdi3.o memcpy.o memset.o \
         strlen.o checksum.o blockops.o memscan.o memcmp.o strncmp.o \
 	strncpy_from_user.o divdi3.o udivdi3.o strlen_user.o \
-	copy_user.o locks.o atomic.o bitops.o
+	copy_user.o locks.o atomic.o bitops.o debuglocks.o
 
 ifdef SMP
 OBJS += irqlock.o
diff --git a/arch/sparc/lib/blockops.S b/arch/sparc/lib/blockops.S
index f8a9e80df..c11ab1b20 100644
--- a/arch/sparc/lib/blockops.S
+++ b/arch/sparc/lib/blockops.S
@@ -1,4 +1,4 @@
-/* $Id: blockops.S,v 1.5 1996/09/24 05:22:56 davem Exp $
+/* $Id: blockops.S,v 1.6 1997/05/03 02:01:54 davem Exp $
  * blockops.S: Common block zero optimized routines.
  *
  * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
@@ -46,25 +46,7 @@
 	.text
 	.align	4
 
-	.globl	C_LABEL(bzero_2page), C_LABEL(bzero_1page)
-C_LABEL(bzero_2page):
-	/* %o0 = buf */
-	or	%g0, %g0, %g1
-	or	%o0, %g0, %o1
-	or	%g0, 0x20, %g2
-1:
-	BLAST_BLOCK(%o0, 0x00)
-	BLAST_BLOCK(%o0, 0x40)
-	BLAST_BLOCK(%o0, 0x80)
-	BLAST_BLOCK(%o0, 0xc0)
-	subcc	%g2, 1, %g2
-	bne	1b
-	 add	%o0, 0x100, %o0
-
-	retl
-	 mov	%o1, %o0
-
-C_LABEL(bzero_1page):
+generic_bzero_1page:
 	/* %o0 = buf */
 	or	%g0, %g0, %g1
 	or	%o0, %g0, %o1
@@ -79,10 +61,9 @@ C_LABEL(bzero_1page):
 	 add	%o0, 0x100, %o0
 
 	retl
-	 mov	%o1, %o0
+	 nop
 
-	.globl	C_LABEL(__copy_1page)
-C_LABEL(__copy_1page):
+__generic_copy_1page:
 	/* %o0 = dst, %o1 = src */
 	or	%g0, 0x10, %g1
 1:
@@ -101,3 +82,9 @@ C_LABEL(__copy_1page):
 
 	retl
 	 nop
+
+	.data
+	.align 4
+	.globl	C_LABEL(bzero_1page), C_LABEL(__copy_1page)
+C_LABEL(bzero_1page):	.word	generic_bzero_1page
+C_LABEL(__copy_1page):	.word	__generic_copy_1page
diff --git a/arch/sparc/lib/debuglocks.c b/arch/sparc/lib/debuglocks.c
new file mode 100644
index 000000000..006cba5a8
--- /dev/null
+++ b/arch/sparc/lib/debuglocks.c
@@ -0,0 +1,463 @@
+/* $Id: debuglocks.c,v 1.1 1997/05/08 18:13:34 davem Exp $
+ * debuglocks.c: Debugging versions of SMP locking primitives.
+ *
+ * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
+ */
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <asm/psr.h>
+#include <asm/system.h>
+#include <asm/spinlock.h>
+
+/* To enable this code, just define SPIN_LOCK_DEBUG in asm/spinlock.h */
+#ifdef SPIN_LOCK_DEBUG
+
+/* Some notes on how these debugging routines work.  When a lock is acquired
+ * an extra debugging member lock->owner_pc is set to the caller of the lock
+ * acquisition routine.  Right before releasing a lock, the debugging program
+ * counter is cleared to zero.
+ *
+ * Furthermore, since PC's are 4 byte aligned on Sparc, we stuff the CPU
+ * number of the owner in the lowest two bits.
+ */
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("spin_lock(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", lock, cpu, caller, lock->owner_pc & ~3, lock->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _spin_lock(spinlock_t *lock)
+{
+	unsigned long caller;
+	unsigned long val;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+again:
+	__asm__ __volatile__("ldstub [%1], %0" : "=r" (val) : "r" (&(lock->lock)));
+	if(val) {
+		while(lock->lock) {
+			STUCK;
+			barrier();
+		}
+		goto again;
+	}
+	lock->owner_pc = (cpu & 3) | (caller & ~3);
+}
+
+int _spin_trylock(spinlock_t *lock)
+{
+	unsigned long val;
+	unsigned long caller;
+	int cpu = smp_processor_id();
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+	__asm__ __volatile__("ldstub [%1], %0" : "=r" (val) : "r" (&(lock->lock)));
+	if(!val) {
+		/* We got it, record our identity for debugging. */
+		lock->owner_pc = (cpu & 3) | (caller & ~3);
+	}
+	return val == 0;
+}
+
+void _spin_unlock(spinlock_t *lock)
+{
+	lock->owner_pc = 0;
+	__asm__ __volatile__("stb %%g0, [%0]" : : "r" (&(lock->lock)) : "memory");
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("spin_lock_irq(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", lock, cpu, caller, lock->owner_pc & ~3, lock->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _spin_lock_irq(spinlock_t *lock)
+{
+	unsigned long caller;
+	unsigned long val;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__cli();
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+again:
+	__asm__ __volatile__("ldstub [%1], %0" : "=r" (val) : "r" (&(lock->lock)));
+	if(val) {
+		while(lock->lock) {
+			STUCK;
+			barrier();
+		}
+		goto again;
+	}
+	lock->owner_pc = (cpu & 3) | (caller & ~3);
+}
+
+void _spin_unlock_irq(spinlock_t *lock)
+{
+	lock->owner_pc = 0;
+	__asm__ __volatile__("stb %%g0, [%0]" : : "r" (&(lock->lock)) : "memory");
+	__sti();
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("spin_lock_irq(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", lock, cpu, caller, lock->owner_pc & ~3, lock->owner_pc & 3); stuck = INIT_STUCK; }
+
+/* Caller macro does __save_and_cli(flags) for us. */
+void _spin_lock_irqsave(spinlock_t *lock)
+{
+	unsigned long caller;
+	unsigned long val;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+again:
+	__asm__ __volatile__("ldstub [%1], %0" : "=r" (val) : "r" (&(lock->lock)));
+	if(val) {
+		while(lock->lock) {
+			STUCK;
+			barrier();
+		}
+		goto again;
+	}
+	lock->owner_pc = (cpu & 3) | (caller & ~3);
+}
+
+void _spin_unlock_irqrestore(spinlock_t *lock)
+{
+	lock->owner_pc = 0;
+	__asm__ __volatile__("stb %%g0, [%0]" : : "r" (&(lock->lock)) : "memory");
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("read_lock(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _read_lock(rwlock_t *rw)
+{
+	unsigned long flags;
+	unsigned long caller;
+	unsigned long val;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+	__save_and_cli(flags);
+wlock_again:
+	__asm__ __volatile__("ldstub [%1 + 3], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff) {
+			STUCK;
+			barrier();
+		}
+		goto wlock_again;
+	}
+clock_again:
+	__asm__ __volatile__("ldstub [%1 + 2], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff00) {
+			STUCK;
+			barrier();
+		}
+		goto clock_again;
+	}
+	(*((unsigned short *)&rw->lock))++;
+	barrier();
+	(*(((unsigned short *)&rw->lock)+1)) = 0;
+	__restore_flags(flags);
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("read_unlock(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _read_unlock(rwlock_t *rw)
+{
+	unsigned long flags, val, caller;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+	__save_and_cli(flags);
+clock_again:
+	__asm__ __volatile__("ldstub [%1 + 2], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff00) {
+			STUCK;
+			barrier();
+		}
+		goto clock_again;
+	}
+	(*((unsigned short *)&rw->lock))--;
+	barrier();
+	(*(((unsigned char *)&rw->lock)+2))=0;
+	__restore_flags(flags);
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("write_lock(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _write_lock(rwlock_t *rw)
+{
+	unsigned long flags, val, caller;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+	__save_and_cli(flags);
+wlock_again:
+	__asm__ __volatile__("ldstub [%1 + 3], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff) {
+			STUCK;
+			barrier();
+		}
+		goto wlock_again;
+	}
+	rw->owner_pc = (cpu & 3) | (caller & ~3);
+	while(rw->lock & ~0xff) {
+		STUCK;
+		barrier();
+	}
+}
+
+void _write_unlock(rwlock_t *rw)
+{
+	rw->owner_pc = 0;
+	barrier();
+	rw->lock = 0;
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("read_lock_irq(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _read_lock_irq(rwlock_t *rw)
+{
+	unsigned long caller;
+	unsigned long val;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+	__cli();
+wlock_again:
+	__asm__ __volatile__("ldstub [%1 + 3], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff) {
+			STUCK;
+			barrier();
+		}
+		goto wlock_again;
+	}
+clock_again:
+	__asm__ __volatile__("ldstub [%1 + 2], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff00) {
+			STUCK;
+			barrier();
+		}
+		goto clock_again;
+	}
+	(*((unsigned short *)&rw->lock))++;
+	barrier();
+	(*(((unsigned short *)&rw->lock)+1)) = 0;
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("read_unlock_irq(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _read_unlock_irq(rwlock_t *rw)
+{
+	unsigned long val, caller;
+	int stuck = INIT_STUCK;
+	int cpu = smp_processor_id();
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+clock_again:
+	__asm__ __volatile__("ldstub [%1 + 2], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff00) {
+			STUCK;
+			barrier();
+		}
+		goto clock_again;
+	}
+	(*((unsigned short *)&rw->lock))--;
+	barrier();
+	(*(((unsigned char *)&rw->lock)+2))=0;
+	__sti();
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("write_lock_irq(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _write_lock_irq(rwlock_t *rw)
+{
+	unsigned long val, caller;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+	__cli();
+wlock_again:
+	__asm__ __volatile__("ldstub [%1 + 3], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff) {
+			STUCK;
+			barrier();
+		}
+		goto wlock_again;
+	}
+	rw->owner_pc = (cpu & 3) | (caller & ~3);
+	while(rw->lock & ~0xff) {
+		STUCK;
+		barrier();
+	}
+}
+
+void _write_unlock_irq(rwlock_t *rw)
+{
+	rw->owner_pc = 0;
+	barrier();
+	rw->lock = 0;
+	__sti();
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("read_lock_irqsave(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+/* Caller does __save_and_cli(flags) for us. */
+void _read_lock_irqsave(rwlock_t *rw)
+{
+	unsigned long caller;
+	unsigned long val;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+wlock_again:
+	__asm__ __volatile__("ldstub [%1 + 3], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff) {
+			STUCK;
+			barrier();
+		}
+		goto wlock_again;
+	}
+clock_again:
+	__asm__ __volatile__("ldstub [%1 + 2], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff00) {
+			STUCK;
+			barrier();
+		}
+		goto clock_again;
+	}
+	(*((unsigned short *)&rw->lock))++;
+	barrier();
+	(*(((unsigned short *)&rw->lock)+1)) = 0;
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("read_unlock_irqrestore(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+void _read_unlock_irqrestore(rwlock_t *rw)
+{
+	unsigned long val, caller;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+clock_again:
+	__asm__ __volatile__("ldstub [%1 + 2], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff00) {
+			STUCK;
+			barrier();
+		}
+		goto clock_again;
+	}
+	(*((unsigned short *)&rw->lock))--;
+	barrier();
+	(*(((unsigned char *)&rw->lock)+2))=0;
+}
+
+#undef INIT_STUCK
+#define INIT_STUCK 100000000
+
+#undef STUCK
+#define STUCK \
+if(!--stuck) { printk("write_lock_irqsave(%p) CPU#%d stuck at %08lx, owner PC(%08lx):CPU(%lx)\n", rw, cpu, caller, rw->owner_pc & ~3, rw->owner_pc & 3); stuck = INIT_STUCK; }
+
+/* Caller does __save_and_cli(flags) for us. */
+void _write_lock_irqsave(rwlock_t *rw)
+{
+	unsigned long val, caller;
+	int cpu = smp_processor_id();
+	int stuck = INIT_STUCK;
+
+	__asm__ __volatile__("mov %%i7, %0" : "=r" (caller));
+wlock_again:
+	__asm__ __volatile__("ldstub [%1 + 3], %0" : "=r" (val) : "r" (&(rw->lock)));
+	if(val) {
+		while(rw->lock & 0xff) {
+			STUCK;
+			barrier();
+		}
+		goto wlock_again;
+	}
+	rw->owner_pc = (cpu & 3) | (caller & ~3);
+	while(rw->lock & ~0xff) {
+		STUCK;
+		barrier();
+	}
+}
+
+void _write_unlock_irqrestore(rwlock_t *rw)
+{
+	rw->owner_pc = 0;
+	barrier();
+	rw->lock = 0;
+}
+
+#endif /* SPIN_LOCK_DEBUG */
diff --git a/arch/sparc/lib/irqlock.S b/arch/sparc/lib/irqlock.S
index db7dd26c3..23194e723 100644
--- a/arch/sparc/lib/irqlock.S
+++ b/arch/sparc/lib/irqlock.S
@@ -1,4 +1,4 @@
-/* $Id: irqlock.S,v 1.2 1997/04/19 04:33:37 davem Exp $
+/* $Id: irqlock.S,v 1.4 1997/05/01 02:26:54 davem Exp $
  * irqlock.S: High performance IRQ global locking and interrupt entry.
  *
  * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
diff --git a/arch/sparc/mm/Makefile b/arch/sparc/mm/Makefile
index 4ae57f18f..f7b9b367c 100644
--- a/arch/sparc/mm/Makefile
+++ b/arch/sparc/mm/Makefile
@@ -1,4 +1,4 @@
-# $Id: Makefile,v 1.24 1997/04/20 14:11:49 ecd Exp $
+# $Id: Makefile,v 1.25 1997/05/03 05:09:11 davem Exp $
 # Makefile for the linux Sparc-specific parts of the memory manager.
 #
 # Note! Dependencies are done automagically by 'make dep', which also
@@ -9,12 +9,30 @@
 
 O_TARGET := mm.o
 O_OBJS   := fault.o init.o sun4c.o srmmu.o hypersparc.o viking.o \
-	    loadmmu.o generic.o asyncd.o extable.o
+	    tsunami.o loadmmu.o generic.o asyncd.o extable.o
 
 include $(TOPDIR)/Rules.make
 
+ifdef SMP
+
+hypersparc.o: hypersparc.S
+	$(CC) -D__ASSEMBLY__ $(AFLAGS) -ansi -c -o hypersparc.o hypersparc.S
+
+viking.o: viking.S
+	$(CC) -D__ASSEMBLY__ $(AFLAGS) -ansi -c -o viking.o viking.S
+
+tsunami.o: tsunami.S
+	$(CC) -D__ASSEMBLY__ $(AFLAGS) -ansi -c -o tsunami.o tsunami.S
+
+else
+
 hypersparc.o: hypersparc.S
 	$(CC) -D__ASSEMBLY__ -ansi -c -o hypersparc.o hypersparc.S
 
 viking.o: viking.S
 	$(CC) -D__ASSEMBLY__ -ansi -c -o viking.o viking.S
+
+tsunami.o: tsunami.S
+	$(CC) -D__ASSEMBLY__ -ansi -c -o tsunami.o tsunami.S
+
+endif
diff --git a/arch/sparc/mm/asyncd.c b/arch/sparc/mm/asyncd.c
index 5d9d476a5..46635db97 100644
--- a/arch/sparc/mm/asyncd.c
+++ b/arch/sparc/mm/asyncd.c
@@ -1,4 +1,4 @@
-/*  $Id: asyncd.c,v 1.9 1996/12/18 06:43:22 tridge Exp $
+/*  $Id: asyncd.c,v 1.10 1997/05/15 21:14:24 davem Exp $
  *  The asyncd kernel daemon. This handles paging on behalf of 
  *  processes that receive page faults due to remote (async) memory
  *  accesses. 
@@ -153,7 +153,7 @@ static int fault_in_page(int taskid,
 	if(!pte)
 		goto no_memory;
 	if(!pte_present(*pte)) {
-		do_no_page(tsk, vma, address, write);
+		handle_mm_fault(tsk, vma, address, write);
 		goto finish_up;
 	}
 	set_pte(pte, pte_mkyoung(*pte));
@@ -165,12 +165,11 @@ static int fault_in_page(int taskid,
 		flush_tlb_page(vma, address);
 		goto finish_up;
 	}
-	do_wp_page(tsk, vma, address, write);
+	handle_mm_fault(tsk, vma, address, write);
 
 	/* Fall through for do_wp_page */
 finish_up:
 	stats.success++;
-	update_mmu_cache(vma, address, *pte);
 	return 0;
 
 no_memory:
diff --git a/arch/sparc/mm/fault.c b/arch/sparc/mm/fault.c
index cfac6bcc2..0d6490860 100644
--- a/arch/sparc/mm/fault.c
+++ b/arch/sparc/mm/fault.c
@@ -1,4 +1,4 @@
-/* $Id: fault.c,v 1.91 1997/03/18 17:56:00 jj Exp $
+/* $Id: fault.c,v 1.92 1997/05/15 21:14:21 davem Exp $
  * fault.c:  Page fault handlers for the Sparc.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -229,7 +229,7 @@ good_area:
 		if(!(vma->vm_flags & (VM_READ | VM_EXEC)))
 			goto bad_area;
 	}
-	handle_mm_fault(vma, address, write);
+	handle_mm_fault(current, vma, address, write);
 	up(&mm->mmap_sem);
 	goto out;
 	/*
@@ -370,7 +370,7 @@ good_area:
 	else
 		if(!(vma->vm_flags & (VM_READ | VM_EXEC)))
 			goto bad_area;
-	handle_mm_fault(vma, address, write);
+	handle_mm_fault(current, vma, address, write);
 	up(&mm->mmap_sem);
 	return;
 bad_area:
diff --git a/arch/sparc/mm/hypersparc.S b/arch/sparc/mm/hypersparc.S
index 4e5a19301..62e2022e0 100644
--- a/arch/sparc/mm/hypersparc.S
+++ b/arch/sparc/mm/hypersparc.S
@@ -1,4 +1,4 @@
-/* $Id: hypersparc.S,v 1.4 1997/04/19 04:33:39 davem Exp $
+/* $Id: hypersparc.S,v 1.7 1997/05/03 05:09:12 davem Exp $
  * hypersparc.S: High speed Hypersparc mmu/cache operations.
  *
  * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
@@ -79,7 +79,7 @@ hypersparc_flush_cache_mm:
 	 sta	%g0, [%o0 + %o4] ASI_M_FLUSH_USER
 hypersparc_flush_cache_mm_out:
 	retl
-	 sta	%g0, [%g0 + %g0] ASI_M_FLUSH_IWHOLE	! hyper_flush_whole_icache
+	 nop
 
 	/* The things we do for performance... */
 hypersparc_flush_cache_range:
@@ -126,7 +126,7 @@ hypersparc_flush_cache_range:
 	bne	1b
 	 sta	%g0, [%o3 + %g5] ASI_M_FLUSH_USER
 	retl
-	 sta	%g0, [%g0 + %g0] ASI_M_FLUSH_IWHOLE
+	 nop
 
 	/* Below our threshold, flush one page at a time. */
 0:
@@ -166,7 +166,7 @@ hypersparc_flush_cache_range:
 	sta	%o3, [%g7] ASI_M_MMUREGS
 hypersparc_flush_cache_range_out:
 	retl
-	 sta	%g0, [%g0 + %g0] ASI_M_FLUSH_IWHOLE
+	 nop
 
 	/* HyperSparc requires a valid mapping where we are about to flush
 	 * in order to check for a physical tag match during the flush.
@@ -221,12 +221,12 @@ hypersparc_flush_cache_page:
 	sta	%o2, [%g4] ASI_M_MMUREGS
 hypersparc_flush_cache_page_out:
 	retl
-	 sta	%g0, [%g0 + %g0] ASI_M_FLUSH_IWHOLE
+	 nop
 
 hypersparc_flush_sig_insns:
-	flush	%o2
+	flush	%o1
 	retl
-	 flush	%o2 + 4
+	 flush	%o1 + 4
 
 	/* HyperSparc is copy-back. */
 hypersparc_flush_page_to_ram:
@@ -289,7 +289,7 @@ hypersparc_flush_tlb_mm:
 	cmp	%o1, -1
 	be	hypersparc_flush_tlb_mm_out
 #endif
-	mov	0x300, %g2
+	 mov	0x300, %g2
 	sta	%o1, [%g1] ASI_M_MMUREGS
 	sta	%g0, [%g2] ASI_M_FLUSH_PROBE
 hypersparc_flush_tlb_mm_out:
@@ -304,7 +304,7 @@ hypersparc_flush_tlb_range:
 	cmp	%o3, -1
 	be	hypersparc_flush_tlb_range_out
 #endif
-	srl	%o1, SRMMU_PGDIR_SHIFT, %o1
+	 srl	%o1, SRMMU_PGDIR_SHIFT, %o1
 	sta	%o3, [%g1] ASI_M_MMUREGS
 	sll	%o1, SRMMU_PGDIR_SHIFT, %o1
 	sethi	%hi(1 << SRMMU_PGDIR_SHIFT), %o4
@@ -324,13 +324,67 @@ hypersparc_flush_tlb_page:
 	mov	SRMMU_CTX_REG, %g1
 	ld	[%o0 + AOFF_mm_context], %o3
 	andn	%o1, (PAGE_SIZE - 1), %o1
-	lda	[%g1] ASI_M_MMUREGS, %g5
 #ifndef __SMP__
 	cmp	%o3, -1
 	be	hypersparc_flush_tlb_page_out
 #endif
+	 lda	[%g1] ASI_M_MMUREGS, %g5
 	sta	%o3, [%g1] ASI_M_MMUREGS
 	sta	%g0, [%o1] ASI_M_FLUSH_PROBE
 hypersparc_flush_tlb_page_out:
 	retl
 	 sta	%g5, [%g1] ASI_M_MMUREGS
+
+	/* High speed page clear/copy. */
+	.globl	hypersparc_bzero_1page, hypersparc_copy_1page
+hypersparc_bzero_1page:
+	clr	%g1
+	mov	32, %g2
+	add	%g2, %g2, %g3
+	add	%g2, %g3, %g4
+	add	%g2, %g4, %g5
+	add	%g2, %g5, %g7
+	add	%g2, %g7, %o2
+	add	%g2, %o2, %o3
+	mov	16, %o1
+1:
+	stda	%g0, [%o0 + %g0] ASI_M_BFILL
+	stda	%g0, [%o0 + %g2] ASI_M_BFILL
+	stda	%g0, [%o0 + %g3] ASI_M_BFILL
+	stda	%g0, [%o0 + %g4] ASI_M_BFILL
+	stda	%g0, [%o0 + %g5] ASI_M_BFILL
+	stda	%g0, [%o0 + %g7] ASI_M_BFILL
+	stda	%g0, [%o0 + %o2] ASI_M_BFILL
+	stda	%g0, [%o0 + %o3] ASI_M_BFILL
+	subcc	%o1, 1, %o1
+	bne	1b
+	 add	%o0, 256, %o0
+
+	retl
+	 nop
+
+hypersparc_copy_1page:
+	sub	%o1, %o0, %o2		! difference
+	mov	16, %g1
+1:
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	add	%o0, 32, %o0
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	add	%o0, 32, %o0
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	add	%o0, 32, %o0
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	add	%o0, 32, %o0
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	add	%o0, 32, %o0
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	add	%o0, 32, %o0
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	add	%o0, 32, %o0
+	sta	%o0, [%o0 + %o2] ASI_M_BCOPY
+	subcc	%g1, 1, %g1
+	bne	1b
+	 add	%o0, 32, %o0
+
+	retl
+	 nop
diff --git a/arch/sparc/mm/srmmu.c b/arch/sparc/mm/srmmu.c
index 9d3afdbdf..b04064efb 100644
--- a/arch/sparc/mm/srmmu.c
+++ b/arch/sparc/mm/srmmu.c
@@ -1,4 +1,4 @@
-/* $Id: srmmu.c,v 1.136 1997/04/20 14:11:51 ecd Exp $
+/* $Id: srmmu.c,v 1.146 1997/05/18 21:11:09 davem Exp $
  * srmmu.c:  SRMMU specific routines for memory management.
  *
  * Copyright (C) 1995 David S. Miller  (davem@caip.rutgers.edu)
@@ -77,12 +77,20 @@ ctxd_t *srmmu_context_table;
 /* Don't change this without changing access to this
  * in arch/sparc/mm/viking.S
  */
-struct srmmu_trans {
+static struct srmmu_trans {
 	unsigned long vbase;
 	unsigned long pbase;
 	unsigned long size;
 } srmmu_map[SPARC_PHYS_BANKS];
 
+#define SRMMU_HASHSZ	256
+
+/* Not static, viking.S uses it. */
+struct srmmu_trans *srmmu_v2p_hash[SRMMU_HASHSZ];
+static struct srmmu_trans *srmmu_p2v_hash[SRMMU_HASHSZ];
+
+#define srmmu_ahashfn(addr)	((addr) >> 24)
+
 static int viking_mxcc_present = 0;
 
 void srmmu_frob_mem_map(unsigned long start_mem)
@@ -113,31 +121,26 @@ void srmmu_frob_mem_map(unsigned long start_mem)
 
 /* Physical memory can be _very_ non-contiguous on the sun4m, especially
  * the SS10/20 class machines and with the latest openprom revisions.
- * So we have to crunch the free page pool.
+ * So we have to do a quick lookup.
  */
 static inline unsigned long srmmu_v2p(unsigned long vaddr)
 {
-	int i;
+	struct srmmu_trans *tp = srmmu_v2p_hash[srmmu_ahashfn(vaddr)];
 
-	for(i=0; srmmu_map[i].size != 0; i++) {
-		if(srmmu_map[i].vbase <= vaddr &&
-		   (srmmu_map[i].vbase + srmmu_map[i].size > vaddr)) {
-			return (vaddr - srmmu_map[i].vbase) + srmmu_map[i].pbase;
-		}
-	}
-	return 0xffffffffUL;
+	if(tp)
+		return (vaddr - tp->vbase + tp->pbase);
+	else
+		return 0xffffffffUL;
 }
 
 static inline unsigned long srmmu_p2v(unsigned long paddr)
 {
-	int i;
+	struct srmmu_trans *tp = srmmu_p2v_hash[srmmu_ahashfn(paddr)];
 
-	for(i=0; srmmu_map[i].size != 0; i++) {
-		if(srmmu_map[i].pbase <= paddr &&
-		   (srmmu_map[i].pbase + srmmu_map[i].size > paddr))
-			return (paddr - srmmu_map[i].pbase) + srmmu_map[i].vbase;
-	}
-	return 0xffffffffUL;
+	if(tp)
+		return (paddr - tp->pbase + tp->vbase);
+	else
+		return 0xffffffffUL;
 }
 
 /* In general all page table modifications should use the V8 atomic
@@ -659,27 +662,6 @@ static void srmmu_set_pte_cacheable(pte_t *ptep, pte_t pteval)
 	srmmu_set_entry(ptep, pte_val(pteval));
 }
 
-static void srmmu_set_pte_nocache_hyper(pte_t *ptep, pte_t pteval)
-{
-	unsigned long page = ((unsigned long)ptep) & PAGE_MASK;
-
-	srmmu_set_entry(ptep, pte_val(pteval));
-	__asm__ __volatile__("
-	lda	[%0] %2, %%g4
-	orcc	%%g4, 0x0, %%g0
-	be	2f
-	 sethi	%%hi(%7), %%g5
-1:	subcc	%%g5, %6, %%g5		! hyper_flush_cache_page
-	bne	1b
-	 sta	%%g0, [%1 + %%g5] %3
-	lda	[%4] %5, %%g0
-2:"	: /* no outputs */
-	: "r" (page | 0x400), "r" (page), "i" (ASI_M_FLUSH_PROBE),
-	  "i" (ASI_M_FLUSH_PAGE), "r" (SRMMU_FAULT_STATUS), "i" (ASI_M_MMUREGS),
-	  "r" (vac_line_size), "i" (PAGE_SIZE)
-	: "g4", "g5", "cc");
-}
-
 static void srmmu_set_pte_nocache_cypress(pte_t *ptep, pte_t pteval)
 {
 	register unsigned long a, b, c, d, e, f, g;
@@ -860,134 +842,27 @@ static void srmmu_unlockarea(char *vaddr, unsigned long len)
  */
 struct task_struct *srmmu_alloc_task_struct(void)
 {
-	return (struct task_struct *) kmalloc(sizeof(struct task_struct), GFP_KERNEL);
-}
-
-unsigned long srmmu_alloc_kernel_stack(struct task_struct *tsk)
-{
-	unsigned long kstk = __get_free_pages(GFP_KERNEL, 1, 0);
-
-	if(!kstk)
-		kstk = (unsigned long) vmalloc(PAGE_SIZE << 1);
-
-	return kstk;
+	return (struct task_struct *) __get_free_pages(GFP_KERNEL, 1, 0);
 }
 
 static void srmmu_free_task_struct(struct task_struct *tsk)
 {
-	kfree(tsk);
-}
-
-static void srmmu_free_kernel_stack(unsigned long stack)
-{
-	if(stack < VMALLOC_START)
-		free_pages(stack, 1);
-	else
-		vfree((char *)stack);
-}
-
-/* Tsunami flushes.  It's page level tlb invalidation is not very
- * useful at all, you must be in the context that page exists in to
- * get a match.
- */
-static void tsunami_flush_cache_all(void)
-{
-	flush_user_windows();
-	tsunami_flush_icache();
-	tsunami_flush_dcache();
-}
-
-static void tsunami_flush_cache_mm(struct mm_struct *mm)
-{
-	FLUSH_BEGIN(mm)
-	flush_user_windows();
-	tsunami_flush_icache();
-	tsunami_flush_dcache();
-	FLUSH_END
-}
-
-static void tsunami_flush_cache_range(struct mm_struct *mm, unsigned long start, unsigned long end)
-{
-	FLUSH_BEGIN(mm)
-	flush_user_windows();
-	tsunami_flush_icache();
-	tsunami_flush_dcache();
-	FLUSH_END
-}
-
-static void tsunami_flush_cache_page(struct vm_area_struct *vma, unsigned long page)
-{
-	FLUSH_BEGIN(vma->vm_mm)
-	flush_user_windows();
-	tsunami_flush_icache();
-	tsunami_flush_dcache();
-	FLUSH_END
-}
-
-/* Tsunami does not have a Copy-back style virtual cache. */
-static void tsunami_flush_page_to_ram(unsigned long page)
-{
-}
-
-/* However, Tsunami is not IO coherent. */
-static void tsunami_flush_page_for_dma(unsigned long page)
-{
-	tsunami_flush_icache();
-	tsunami_flush_dcache();
-}
-
-/* Tsunami has harvard style split I/D caches which do not snoop each other,
- * so we have to flush on-stack sig insns.  Only the icache need be flushed
- * since the Tsunami has a write-through data cache.
- */
-static void tsunami_flush_sig_insns(struct mm_struct *mm, unsigned long insn_addr)
-{
-	tsunami_flush_icache();
-}
-
-static void tsunami_flush_chunk(unsigned long chunk)
-{
+	free_pages((unsigned long)tsk, 1);
 }
 
-static void tsunami_flush_tlb_all(void)
-{
-	srmmu_flush_whole_tlb();
-	module_stats.invall++;
-}
-
-static void tsunami_flush_tlb_mm(struct mm_struct *mm)
-{
-	FLUSH_BEGIN(mm)
-	srmmu_flush_whole_tlb();
-	module_stats.invmm++;
-	FLUSH_END
-}
-
-static void tsunami_flush_tlb_range(struct mm_struct *mm, unsigned long start, unsigned long end)
-{
-	FLUSH_BEGIN(mm)
-	srmmu_flush_whole_tlb();
-	module_stats.invrnge++;
-	FLUSH_END
-}
-
-static void tsunami_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
-{
-	struct mm_struct *mm = vma->vm_mm;
-
-	FLUSH_BEGIN(mm)
-	__asm__ __volatile__("
-	lda	[%0] %3, %%g5
-	sta	%1, [%0] %3
-	sta	%%g0, [%2] %4
-	sta	%%g5, [%0] %3"
-	: /* no outputs */
-	: "r" (SRMMU_CTX_REG), "r" (mm->context), "r" (page & PAGE_MASK),
-	  "i" (ASI_M_MMUREGS), "i" (ASI_M_FLUSH_PROBE)
-	: "g5");
-	module_stats.invpg++;
-	FLUSH_END
-}
+/* tsunami.S */
+extern void tsunami_flush_cache_all(void);
+extern void tsunami_flush_cache_mm(struct mm_struct *mm);
+extern void tsunami_flush_cache_range(struct mm_struct *mm, unsigned long start, unsigned long end);
+extern void tsunami_flush_cache_page(struct vm_area_struct *vma, unsigned long page);
+extern void tsunami_flush_page_to_ram(unsigned long page);
+extern void tsunami_flush_page_for_dma(unsigned long page);
+extern void tsunami_flush_sig_insns(struct mm_struct *mm, unsigned long insn_addr);
+extern void tsunami_flush_chunk(unsigned long chunk);
+extern void tsunami_flush_tlb_all(void);
+extern void tsunami_flush_tlb_mm(struct mm_struct *mm);
+extern void tsunami_flush_tlb_range(struct mm_struct *mm, unsigned long start, unsigned long end);
+extern void tsunami_flush_tlb_page(struct vm_area_struct *vma, unsigned long page);
 
 /* Swift flushes.  It has the recommended SRMMU specification flushing
  * facilities, so we can do things in a more fine grained fashion than we
@@ -1364,18 +1239,31 @@ extern void hypersparc_flush_tlb_all(void);
 extern void hypersparc_flush_tlb_mm(struct mm_struct *mm);
 extern void hypersparc_flush_tlb_range(struct mm_struct *mm, unsigned long start, unsigned long end);
 extern void hypersparc_flush_tlb_page(struct vm_area_struct *vma, unsigned long page);
+extern void hypersparc_bzero_1page(void *);
+extern void hypersparc_copy_1page(void *, const void *);
+
+static void srmmu_set_pte_nocache_hyper(pte_t *ptep, pte_t pteval)
+{
+	unsigned long page = ((unsigned long)ptep) & PAGE_MASK;
+
+	srmmu_set_entry(ptep, pte_val(pteval));
+	hypersparc_flush_page_to_ram(page);
+}
 
 static void hypersparc_ctxd_set(ctxd_t *ctxp, pgd_t *pgdp)
 {
+	srmmu_set_entry((pte_t *)ctxp, __pte((SRMMU_ET_PTD | (srmmu_v2p((unsigned long) pgdp) >> 4))));
+	hypersparc_flush_page_to_ram((unsigned long)ctxp);
 	hyper_flush_whole_icache();
-	set_pte((pte_t *)ctxp, __pte((SRMMU_ET_PTD | (srmmu_v2p((unsigned long) pgdp) >> 4))));
 }
 
 static void hypersparc_update_rootmmu_dir(struct task_struct *tsk, pgd_t *pgdp) 
 {
 	unsigned long page = ((unsigned long) pgdp) & PAGE_MASK;
 
-	hypersparc_flush_page_to_ram(page);
+	if(pgdp != swapper_pg_dir)
+		hypersparc_flush_page_to_ram(page);
+
 	if(tsk->mm->context != NO_CONTEXT) {
 		flush_cache_mm(tsk->mm);
 		ctxd_set(&srmmu_context_table[tsk->mm->context], pgdp);
@@ -1429,26 +1317,29 @@ static void cypress_update_rootmmu_dir(struct task_struct *tsk, pgd_t *pgdp)
 
 static void hypersparc_switch_to_context(struct task_struct *tsk)
 {
-	hyper_flush_whole_icache();
 	if(tsk->mm->context == NO_CONTEXT) {
+		ctxd_t *ctxp;
+
 		alloc_context(tsk->mm);
-		flush_cache_mm(tsk->mm);
-		ctxd_set(&srmmu_context_table[tsk->mm->context], tsk->mm->pgd);
-		flush_tlb_mm(tsk->mm);
+		ctxp = &srmmu_context_table[tsk->mm->context];
+		srmmu_set_entry((pte_t *)ctxp, __pte((SRMMU_ET_PTD | (srmmu_v2p((unsigned long) tsk->mm->pgd) >> 4))));
+		hypersparc_flush_page_to_ram((unsigned long)ctxp);
 	}
+	hyper_flush_whole_icache();
 	srmmu_set_context(tsk->mm->context);
 }
 
 static void hypersparc_init_new_context(struct mm_struct *mm)
 {
-	hyper_flush_whole_icache();
+	ctxd_t *ctxp;
 
 	alloc_context(mm);
 
-	flush_cache_mm(mm);
-	ctxd_set(&srmmu_context_table[mm->context], mm->pgd);
-	flush_tlb_mm(mm);
+	ctxp = &srmmu_context_table[mm->context];
+	srmmu_set_entry((pte_t *)ctxp, __pte((SRMMU_ET_PTD | (srmmu_v2p((unsigned long) mm->pgd) >> 4))));
+	hypersparc_flush_page_to_ram((unsigned long)ctxp);
 
+	hyper_flush_whole_icache();
 	if(mm == current->mm)
 		srmmu_set_context(mm->context);
 }
@@ -2150,6 +2041,32 @@ check_and_return:
 	MKTRACE(("success\n"));
 	init_task.mm->mmap->vm_start = page_offset = low_base;
 	stack_top = page_offset - PAGE_SIZE;
+#if 1
+	for(entry = 0; srmmu_map[entry].size; entry++) {
+		printk("[%d]: v[%08lx,%08lx](%lx) p[%08lx]\n", entry,
+		       srmmu_map[entry].vbase,
+		       srmmu_map[entry].vbase + srmmu_map[entry].size,
+		       srmmu_map[entry].size,
+		       srmmu_map[entry].pbase);
+	}
+#endif
+
+	/* Now setup the p2v/v2p hash tables. */
+	for(entry = 0; entry < SRMMU_HASHSZ; entry++)
+		srmmu_v2p_hash[entry] = srmmu_p2v_hash[entry] = NULL;
+	for(entry = 0; srmmu_map[entry].size; entry++) {
+		unsigned long addr;
+
+		for(addr = srmmu_map[entry].vbase;
+		    addr < (srmmu_map[entry].vbase + srmmu_map[entry].size);
+		    addr += (1 << 24))
+			srmmu_v2p_hash[srmmu_ahashfn(addr)] = &srmmu_map[entry];
+		for(addr = srmmu_map[entry].pbase;
+		    addr < (srmmu_map[entry].pbase + srmmu_map[entry].size);
+		    addr += (1 << 24))
+			srmmu_p2v_hash[srmmu_ahashfn(addr)] = &srmmu_map[entry];
+	}
+
 	return; /* SUCCESS! */
 }
 
@@ -2338,7 +2255,7 @@ static void srmmu_vac_update_mmu_cache(struct vm_area_struct * vma,
 					start += PAGE_SIZE;
 				}
 			}
-		} while ((vmaring = vmaring->vm_next_share) != inode->i_mmap);
+		} while ((vmaring = vmaring->vm_next_share) != NULL);
 
 		if(alias_found && !(pte_val(pte) & _SUN4C_PAGE_NOCACHE)) {
 			pgdp = srmmu_pgd_offset(vma->vm_mm, address);
@@ -2355,13 +2272,19 @@ static void srmmu_vac_update_mmu_cache(struct vm_area_struct * vma,
 static void hypersparc_destroy_context(struct mm_struct *mm)
 {
 	if(mm->context != NO_CONTEXT && mm->count == 1) {
+		ctxd_t *ctxp;
+
 		/* HyperSparc is copy-back, any data for this
 		 * process in a modified cache line is stale
 		 * and must be written back to main memory now
 		 * else we eat shit later big time.
 		 */
 		flush_cache_mm(mm);
-		ctxd_set(&srmmu_context_table[mm->context], swapper_pg_dir);
+
+		ctxp = &srmmu_context_table[mm->context];
+		srmmu_set_entry((pte_t *)ctxp, __pte((SRMMU_ET_PTD | (srmmu_v2p((unsigned long) swapper_pg_dir) >> 4))));
+		hypersparc_flush_page_to_ram((unsigned long)ctxp);
+
 		flush_tlb_mm(mm);
 		free_context(mm->context);
 		mm->context = NO_CONTEXT;
@@ -2450,6 +2373,11 @@ static void poke_hypersparc(void)
 	hyper_flush_whole_icache();
 	clear = srmmu_get_faddr();
 	clear = srmmu_get_fstatus();
+
+#ifdef __SMP__
+	/* Avoid unnecessary cross calls. */
+	flush_page_for_dma = local_flush_page_for_dma;
+#endif
 }
 
 __initfunc(static void init_hypersparc(void))
@@ -2482,6 +2410,14 @@ __initfunc(static void init_hypersparc(void))
 	update_mmu_cache = srmmu_vac_update_mmu_cache;
 	sparc_update_rootmmu_dir = hypersparc_update_rootmmu_dir;
 	poke_srmmu = poke_hypersparc;
+
+	/* High performance page copy/clear. */
+	{	extern void (*__copy_1page)(void *, const void *);
+		extern void (*bzero_1page)(void *);
+
+		__copy_1page = hypersparc_copy_1page;
+		bzero_1page = hypersparc_bzero_1page;
+	}
 }
 
 static void poke_cypress(void)
@@ -3014,9 +2950,7 @@ __initfunc(void ld_mmu_srmmu(void))
         mmu_p2v = srmmu_p2v;
 
 	/* Task struct and kernel stack allocating/freeing. */
-	alloc_kernel_stack = srmmu_alloc_kernel_stack;
 	alloc_task_struct = srmmu_alloc_task_struct;
-	free_kernel_stack = srmmu_free_kernel_stack;
 	free_task_struct = srmmu_free_task_struct;
 
 	quick_kernel_fault = srmmu_quick_kernel_fault;
diff --git a/arch/sparc/mm/sun4c.c b/arch/sparc/mm/sun4c.c
index ebeada4c7..a0ffc10ed 100644
--- a/arch/sparc/mm/sun4c.c
+++ b/arch/sparc/mm/sun4c.c
@@ -1,4 +1,4 @@
-/* $Id: sun4c.c,v 1.143 1997/04/11 00:42:14 davem Exp $
+/* $Id: sun4c.c,v 1.148 1997/05/18 21:11:19 davem Exp $
  * sun4c.c: Doing in software what should be done in hardware.
  *
  * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
@@ -1093,8 +1093,7 @@ static void sun4c_quick_kernel_fault(unsigned long address)
         panic("sun4c kernel fault handler bolixed...");
 }
 
-/*
- * 4 page buckets for task struct and kernel stack allocation.
+/* 2 page buckets for task struct and kernel stack allocation.
  *
  * TASK_STACK_BEGIN
  * bucket[0]
@@ -1105,24 +1104,17 @@ static void sun4c_quick_kernel_fault(unsigned long address)
  *
  * Each slot looks like:
  *
- *  page 1   --  task struct
- *  page 2   --  unmapped, for stack redzone (maybe use for pgd)
- *  page 3/4 --  kernel stack
+ *  page 1 --  task struct + beginning of kernel stack
+ *  page 2 --  rest of kernel stack
  */
 
-struct task_bucket {
-	struct task_struct task;
-	char _unused1[PAGE_SIZE - sizeof(struct task_struct)];
-	char kstack[(PAGE_SIZE*3)];
-};
-
-struct task_bucket *sun4c_bucket[NR_TASKS];
+union task_union *sun4c_bucket[NR_TASKS];
 
 static int sun4c_lowbucket_avail;
 
-#define BUCKET_EMPTY     ((struct task_bucket *) 0)
-#define BUCKET_SIZE      (PAGE_SIZE << 2)
-#define BUCKET_SHIFT     14        /* log2(sizeof(struct task_bucket)) */
+#define BUCKET_EMPTY     ((union task_union *) 0)
+#define BUCKET_SHIFT     (PAGE_SHIFT + 1)        /* log2(sizeof(struct task_bucket)) */
+#define BUCKET_SIZE      (1 << BUCKET_SHIFT)
 #define BUCKET_NUM(addr) ((((addr) - SUN4C_LOCK_VADDR) >> BUCKET_SHIFT))
 #define BUCKET_ADDR(num) (((num) << BUCKET_SHIFT) + SUN4C_LOCK_VADDR)
 #define BUCKET_PTE(page)       \
@@ -1177,10 +1169,10 @@ static inline void garbage_collect(int entry)
 {
 	int start, end;
 
-	/* 16 buckets per segment... */
-	entry &= ~15;
+	/* 32 buckets per segment... */
+	entry &= ~31;
 	start = entry;
-	for(end = (start + 16); start < end; start++)
+	for(end = (start + 32); start < end; start++)
 		if(sun4c_bucket[start] != BUCKET_EMPTY)
 			return;
 
@@ -1190,121 +1182,70 @@ static inline void garbage_collect(int entry)
 
 static struct task_struct *sun4c_alloc_task_struct(void)
 {
-	unsigned long addr, page;
+	unsigned long addr, pages;
 	int entry;
 
-	page = get_free_page(GFP_KERNEL);
-	if(!page)
+	pages = __get_free_pages(GFP_KERNEL, 1, 0);
+	if(!pages)
 		return (struct task_struct *) 0;
 
 	for(entry = sun4c_lowbucket_avail; entry < NR_TASKS; entry++)
 		if(sun4c_bucket[entry] == BUCKET_EMPTY)
 			break;
 	if(entry == NR_TASKS) {
-		free_page(page);
+		free_pages(pages, 1);
 		return (struct task_struct *) 0;
 	}
 	if(entry >= sun4c_lowbucket_avail)
 		sun4c_lowbucket_avail = entry + 1;
 
 	addr = BUCKET_ADDR(entry);
-	sun4c_bucket[entry] = (struct task_bucket *) addr;
+	sun4c_bucket[entry] = (union task_union *) addr;
 	if(sun4c_get_segmap(addr) == invalid_segment)
 		get_locked_segment(addr);
-	sun4c_put_pte(addr, BUCKET_PTE(page));
+	sun4c_put_pte(addr, BUCKET_PTE(pages));
+	sun4c_put_pte(addr + PAGE_SIZE, BUCKET_PTE(pages + PAGE_SIZE));
 
 	return (struct task_struct *) addr;
 }
 
-static unsigned long sun4c_alloc_kernel_stack(struct task_struct *tsk)
-{
-	unsigned long saddr = (unsigned long) tsk;
-	unsigned long page[2];
-
-	if(!saddr)
-		return 0;
-	page[0] = __get_free_page(GFP_KERNEL);
-	if(!page[0])
-		return 0;
-	page[1] = __get_free_page(GFP_KERNEL);
-	if(!page[1]) {
-		free_page(page[0]);
-		return 0;
-	}
-
-	saddr += PAGE_SIZE << 1;
-	sun4c_put_pte(saddr, BUCKET_PTE(page[0]));
-	sun4c_put_pte(saddr + PAGE_SIZE, BUCKET_PTE(page[1]));
-	return saddr;
-}
-
-static void sun4c_free_kernel_stack_hw(unsigned long stack)
-{
-	unsigned long page[2];
-
-	page[0] = BUCKET_PTE_PAGE(sun4c_get_pte(stack));
-	page[1] = BUCKET_PTE_PAGE(sun4c_get_pte(stack+PAGE_SIZE));
-
-	/* We are deleting a mapping, so the flushes here are mandatory. */
-	sun4c_flush_page_hw(stack);
-	sun4c_flush_page_hw(stack + PAGE_SIZE);
-
-	sun4c_put_pte(stack, 0);
-	sun4c_put_pte(stack + PAGE_SIZE, 0);
-	free_page(page[0]);
-	free_page(page[1]);
-}
-
 static void sun4c_free_task_struct_hw(struct task_struct *tsk)
 {
 	unsigned long tsaddr = (unsigned long) tsk;
-	unsigned long page = BUCKET_PTE_PAGE(sun4c_get_pte(tsaddr));
+	unsigned long pages = BUCKET_PTE_PAGE(sun4c_get_pte(tsaddr));
 	int entry = BUCKET_NUM(tsaddr);
 
 	/* We are deleting a mapping, so the flush here is mandatory. */
 	sun4c_flush_page_hw(tsaddr);
+	sun4c_flush_page_hw(tsaddr + PAGE_SIZE);
 
 	sun4c_put_pte(tsaddr, 0);
+	sun4c_put_pte(tsaddr + PAGE_SIZE, 0);
 	sun4c_bucket[entry] = BUCKET_EMPTY;
 	if(entry < sun4c_lowbucket_avail)
 		sun4c_lowbucket_avail = entry;
 
-	free_page(page);
+	free_pages(pages, 1);
 	garbage_collect(entry);
 }
 
-static void sun4c_free_kernel_stack_sw(unsigned long stack)
-{
-	unsigned long page[2];
-
-	page[0] = BUCKET_PTE_PAGE(sun4c_get_pte(stack));
-	page[1] = BUCKET_PTE_PAGE(sun4c_get_pte(stack+PAGE_SIZE));
-
-	/* We are deleting a mapping, so the flushes here are mandatory. */
-	sun4c_flush_page_sw(stack);
-	sun4c_flush_page_sw(stack + PAGE_SIZE);
-
-	sun4c_put_pte(stack, 0);
-	sun4c_put_pte(stack + PAGE_SIZE, 0);
-	free_page(page[0]);
-	free_page(page[1]);
-}
-
 static void sun4c_free_task_struct_sw(struct task_struct *tsk)
 {
 	unsigned long tsaddr = (unsigned long) tsk;
-	unsigned long page = BUCKET_PTE_PAGE(sun4c_get_pte(tsaddr));
+	unsigned long pages = BUCKET_PTE_PAGE(sun4c_get_pte(tsaddr));
 	int entry = BUCKET_NUM(tsaddr);
 
 	/* We are deleting a mapping, so the flush here is mandatory. */
 	sun4c_flush_page_sw(tsaddr);
+	sun4c_flush_page_sw(tsaddr + PAGE_SIZE);
 
 	sun4c_put_pte(tsaddr, 0);
+	sun4c_put_pte(tsaddr + PAGE_SIZE, 0);
 	sun4c_bucket[entry] = BUCKET_EMPTY;
 	if(entry < sun4c_lowbucket_avail)
 		sun4c_lowbucket_avail = entry;
 
-	free_page(page);
+	free_pages(pages, 1);
 	garbage_collect(entry);
 }
 
@@ -1312,8 +1253,8 @@ __initfunc(static void sun4c_init_buckets(void))
 {
 	int entry;
 
-	if(sizeof(struct task_bucket) != (PAGE_SIZE << 2)) {
-		prom_printf("task bucket not 4 pages!\n");
+	if(sizeof(union task_union) != (PAGE_SIZE << 1)) {
+		prom_printf("task union not 2 pages!\n");
 		prom_halt();
 	}
 	for(entry = 0; entry < NR_TASKS; entry++)
@@ -2526,7 +2467,7 @@ static void sun4c_vac_alias_fixup(struct vm_area_struct *vma, unsigned long addr
 					start += PAGE_SIZE;
 				}
 			}
-		} while ((vmaring = vmaring->vm_next_share) != inode->i_mmap);
+		} while ((vmaring = vmaring->vm_next_share) != NULL);
 
 		if(alias_found && !(pte_val(pte) & _SUN4C_PAGE_NOCACHE)) {
 			pgdp = sun4c_pgd_offset(vma->vm_mm, address);
@@ -2645,7 +2586,6 @@ __initfunc(void ld_mmu_sun4c(void))
 		flush_tlb_mm = sun4c_flush_tlb_mm_hw;
 		flush_tlb_range = sun4c_flush_tlb_range_hw;
 		flush_tlb_page = sun4c_flush_tlb_page_hw;
-		free_kernel_stack = sun4c_free_kernel_stack_hw;
 		free_task_struct = sun4c_free_task_struct_hw;
 		switch_to_context = sun4c_switch_to_context_hw;
 		destroy_context = sun4c_destroy_context_hw;
@@ -2658,7 +2598,6 @@ __initfunc(void ld_mmu_sun4c(void))
 		flush_tlb_mm = sun4c_flush_tlb_mm_sw;
 		flush_tlb_range = sun4c_flush_tlb_range_sw;
 		flush_tlb_page = sun4c_flush_tlb_page_sw;
-		free_kernel_stack = sun4c_free_kernel_stack_sw;
 		free_task_struct = sun4c_free_task_struct_sw;
 		switch_to_context = sun4c_switch_to_context_sw;
 		destroy_context = sun4c_destroy_context_sw;
@@ -2736,7 +2675,6 @@ __initfunc(void ld_mmu_sun4c(void))
         mmu_p2v = sun4c_p2v;
 	
 	/* Task struct and kernel stack allocating/freeing. */
-	alloc_kernel_stack = sun4c_alloc_kernel_stack;
 	alloc_task_struct = sun4c_alloc_task_struct;
 
 	quick_kernel_fault = sun4c_quick_kernel_fault;
diff --git a/arch/sparc/mm/tsunami.S b/arch/sparc/mm/tsunami.S
new file mode 100644
index 000000000..2a598cd7b
--- /dev/null
+++ b/arch/sparc/mm/tsunami.S
@@ -0,0 +1,90 @@
+/* $Id: tsunami.S,v 1.1 1997/05/03 05:09:09 davem Exp $
+ * tsunami.S: High speed MicroSparc-I mmu/cache operations.
+ *
+ * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
+ */
+
+#include <asm/ptrace.h>
+#include <asm/psr.h>
+#include <asm/asi.h>
+#include <asm/page.h>
+#include <asm/pgtsrmmu.h>
+
+#define WINDOW_FLUSH(tmp1, tmp2)					\
+	mov	0, tmp1;						\
+98:	ld	[%g6 + AOFF_task_tss + AOFF_thread_uwinmask], tmp2;	\
+	orcc	%g0, tmp2, %g0;						\
+	add	tmp1, 1, tmp1;						\
+	bne	98b;							\
+	 save	%sp, -64, %sp;						\
+99:	subcc	tmp1, 1, tmp1;						\
+	bne	99b;							\
+	 restore %g0, %g0, %g0;
+
+	.text
+	.align	4
+
+	.globl	tsunami_flush_cache_all, tsunami_flush_cache_mm
+	.globl	tsunami_flush_cache_range, tsunami_flush_cache_page
+	.globl	tsunami_flush_page_to_ram, tsunami_flush_page_for_dma
+	.globl	tsunami_flush_sig_insns, tsunami_flush_chunk
+	.globl	tsunami_flush_tlb_all, tsunami_flush_tlb_mm
+	.globl	tsunami_flush_tlb_range, tsunami_flush_tlb_page
+
+	/* Sliiick... */
+tsunami_flush_cache_page:
+	ld	[%o0 + 0x0], %o0	/* XXX vma->vm_mm, GROSS XXX */
+tsunami_flush_cache_mm:
+tsunami_flush_cache_range:
+	ld	[%o0 + AOFF_mm_context], %g2
+#ifndef __SMP__
+	cmp	%g2, -1
+	be	tsunami_flush_cache_out
+#endif
+tsunami_flush_cache_all:
+	WINDOW_FLUSH(%g4, %g5)
+tsunami_flush_page_for_dma:
+	sta	%g0, [%g0] ASI_M_DC_FLCLEAR
+	sta	%g0, [%g0] ASI_M_IC_FLCLEAR
+tsunami_flush_cache_out:
+tsunami_flush_page_to_ram:
+tsunami_flush_chunk:
+	retl
+	 nop
+
+tsunami_flush_sig_insns:
+	flush	%o1
+	retl
+	 flush	%o1 + 4
+
+	/* More slick stuff... */
+tsunami_flush_tlb_mm:
+tsunami_flush_tlb_range:
+#ifndef __SMP__
+	ld	[%o0 + AOFF_mm_context], %g2
+	cmp	%g2, -1
+	be	tsunami_flush_tlb_out
+#endif
+tsunami_flush_tlb_all:
+	 mov	0x400, %o1
+	sta	%g0, [%o1] ASI_M_FLUSH_PROBE
+tsunami_flush_tlb_out:
+	retl
+	 nop
+
+	/* This one can be done in a fine grained manner... */
+tsunami_flush_tlb_page:
+	ld	[%o0 + 0x00], %o0	/* XXX vma->vm_mm GROSS XXX */
+	mov	SRMMU_CTX_REG, %g1
+	ld	[%o0 + AOFF_mm_context], %o3
+	andn	%o1, (PAGE_SIZE - 1), %o1
+#ifndef __SMP__
+	cmp	%o3, -1
+	be	tsunami_flush_tlb_page_out
+#endif
+	 lda	[%g1] ASI_M_MMUREGS, %g5
+	sta	%o3, [%g1] ASI_M_MMUREGS
+	sta	%g0, [%o1] ASI_M_FLUSH_PROBE
+tsunami_flush_tlb_page_out:
+	retl
+	 sta	%g5, [%g1] ASI_M_MMUREGS
diff --git a/arch/sparc/mm/viking.S b/arch/sparc/mm/viking.S
index f61aa4398..19d426ec7 100644
--- a/arch/sparc/mm/viking.S
+++ b/arch/sparc/mm/viking.S
@@ -1,4 +1,4 @@
-/* $Id: viking.S,v 1.2 1997/04/20 21:21:49 ecd Exp $
+/* $Id: viking.S,v 1.3 1997/05/04 10:02:14 ecd Exp $
  * viking.S: High speed Viking cache/mmu operations
  *
  * Copyright (C) 1997  Eddie C. Dost  (ecd@skynet.be)
@@ -38,40 +38,26 @@
 
 viking_flush_page:
 viking_flush_chunk:
-	sethi	%hi(C_LABEL(srmmu_map)), %g2
-	or	%g2, %lo(C_LABEL(srmmu_map)), %g3
-	ld	[%g3 + 8], %g2
-	cmp	%g2, 0
-	be	3f
+	sethi	%hi(C_LABEL(srmmu_v2p_hash)), %g2
+	or	%g2, %lo(C_LABEL(srmmu_v2p_hash)), %g2
+	srl	%o0, 24, %o1
+	sll	%o1, 2, %o1
+
+	ld	[%g2 + %o1], %g3
+	cmp	%g3, 0
+	bne	1f
 	 and	%o0, PAGE_MASK, %o0
 
-	ld	[%g3], %o1
-1:
-	cmp	%o1, %o0
-	bgu,a	2f
-	 add	%g3, 0xc, %g3
-
-	add	%o1, %g2, %g2
-	cmp	%g2, %o0
-	bleu,a	2f
-	 add	%g3, 0xc, %g3
+	retl
+	 nop
 
+1:
+	ld	[%g3], %o1
 	sub	%o0, %o1, %g2
 	ld	[%g3 + 4], %o0
 	add	%g2, %o0, %g3
-	b	4f
-	 srl	%g3, 12, %g1		! ppage >> 12
-
-2:
-	ld	[%g3 + 8], %g2
-	cmp	%g2, 0
-	bne,a	1b
-	 ld	[%g3], %o1
-3:
-	retl
-	 nop
+	srl	%g3, 12, %g1		! ppage >> 12
 
-4:
 	clr	%o1			! set counter, 0 - 127
 	sethi	%hi(KERNBASE + PAGE_SIZE - 0x80000000), %o3
 	sethi	%hi(0x80000000), %o4
@@ -131,40 +117,27 @@ viking_flush_chunk:
 
 
 viking_mxcc_flush_page:
-	sethi	%hi(C_LABEL(srmmu_map)), %g2
-	or	%g2, %lo(C_LABEL(srmmu_map)), %g3
-	ld	[%g3 + 8], %g2
-	cmp	%g2, 0
-	be	3f
+	sethi	%hi(C_LABEL(srmmu_v2p_hash)), %g2
+	or	%g2, %lo(C_LABEL(srmmu_v2p_hash)), %g2
+	srl	%o0, 24, %o1
+	sll	%o1, 2, %o1
+
+	ld	[%g2 + %o1], %g3
+	cmp	%g3, 0
+	bne	1f
 	 and	%o0, PAGE_MASK, %o0
 
-	ld	[%g3], %o1
-1:
-	cmp	%o1, %o0
-	bgu,a	2f
-	 add	%g3, 0xc, %g3
-
-	add	%o1, %g2, %g2
-	cmp	%g2, %o0
-	bleu,a	2f
-	 add	%g3, 0xc, %g3
+	retl
+	 nop
 
+1:
+	ld	[%g3], %o1
 	sub	%o0, %o1, %g2
 	ld	[%g3 + 4], %o0
+	sethi   %hi(PAGE_SIZE), %g4
 	add	%g2, %o0, %g3
-	sethi	%hi(PAGE_SIZE), %g4
-	b	4f
-	 add	%g3, %g4, %g3			! ppage + PAGE_SIZE
-
-2:
-	ld	[%g3 + 8], %g2
-	cmp	%g2, 0
-	bne,a	1b
-	 ld	[%g3], %o1
-3:
-	retl
-	 nop
-4:
+	add	%g3, %g4, %g3			! ppage + PAGE_SIZE
+
 	mov	0x10, %g2			! set cacheable bit
 	sethi	%hi(MXCC_SRCSTREAM), %o2
 	or	%o2, %lo(MXCC_SRCSTREAM), %o2
diff --git a/arch/sparc/prom/console.c b/arch/sparc/prom/console.c
index 9551094d0..4c999477b 100644
--- a/arch/sparc/prom/console.c
+++ b/arch/sparc/prom/console.c
@@ -1,4 +1,4 @@
-/* $Id: console.c,v 1.11 1997/03/18 17:58:10 jj Exp $
+/* $Id: console.c,v 1.14 1997/05/14 20:44:58 davem Exp $
  * console.c: Routines that deal with sending and receiving IO
  *            to/from the current console device using the PROM.
  *
@@ -14,6 +14,9 @@
 #include <asm/system.h>
 #include <linux/string.h>
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
+
 /* Non blocking get character from console input device, returns -1
  * if no input was taken.  This can be used for polling.
  */
@@ -38,11 +41,12 @@ prom_nbgetchar(void)
 		}
 		break;
 	case PROM_AP1000:
+	default:
 		i = -1;
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	return i; /* Ugh, we could spin forever on unsupported proms ;( */
@@ -81,9 +85,12 @@ prom_nbputchar(char c)
 #endif
 	
 		break;
+	default:
+		i = -1;
+		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	return i; /* Ugh, we could spin forever on unsupported proms ;( */
@@ -130,7 +137,7 @@ prom_query_input_device()
 		save_flags(flags); cli();
 		st_p = (*romvec->pv_v2devops.v2_inst2pkg)(*romvec->pv_v2bootargs.fd_stdin);
 		__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-				     "r" (&current_set[smp_processor_id()]) :
+				     "r" (&current_set[hard_smp_processor_id()]) :
 				     "memory");
 		restore_flags(flags);
 		if(prom_node_has_property(st_p, "keyboard"))
@@ -177,7 +184,7 @@ prom_query_output_device()
 		save_flags(flags); cli();
 		st_p = (*romvec->pv_v2devops.v2_inst2pkg)(*romvec->pv_v2bootargs.fd_stdout);
 		__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-				     "r" (&current_set[smp_processor_id()]) :
+				     "r" (&current_set[hard_smp_processor_id()]) :
 				     "memory");
 		restore_flags(flags);
 		propl = prom_getproperty(st_p, "device_type", propb, sizeof(propb));
@@ -208,6 +215,7 @@ prom_query_output_device()
 		}
 		break;
 	case PROM_AP1000:
+	default:
 		return PROMDEV_I_UNK;
 	};
 	return PROMDEV_O_UNK;
diff --git a/arch/sparc/prom/devmap.c b/arch/sparc/prom/devmap.c
index 841d63a47..cd99ac3d6 100644
--- a/arch/sparc/prom/devmap.c
+++ b/arch/sparc/prom/devmap.c
@@ -1,4 +1,4 @@
-/* $Id: devmap.c,v 1.3 1996/09/19 20:27:19 davem Exp $
+/* $Id: devmap.c,v 1.5 1997/05/14 20:44:59 davem Exp $
  * promdevmap.c:  Map device/IO areas to virtual addresses.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -11,6 +11,9 @@
 #include <asm/openprom.h>
 #include <asm/oplib.h>
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
+
 /* Just like the routines in palloc.c, these should not be used
  * by the kernel at all.  Bootloader facility mainly.  And again,
  * this is only available on V2 proms and above.
@@ -33,7 +36,7 @@ prom_mapio(char *vhint, int ios, unsigned int paddr, unsigned int num_bytes)
 	ret = (*(romvec->pv_v2devops.v2_dumb_mmap))(vhint, ios, paddr,
 						    num_bytes);
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	return ret;
@@ -49,7 +52,7 @@ prom_unmapio(char *vaddr, unsigned int num_bytes)
 	save_flags(flags); cli();
 	(*(romvec->pv_v2devops.v2_dumb_munmap))(vaddr, num_bytes);
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	return;
diff --git a/arch/sparc/prom/devops.c b/arch/sparc/prom/devops.c
index 2b33dd2b7..f7feb0815 100644
--- a/arch/sparc/prom/devops.c
+++ b/arch/sparc/prom/devops.c
@@ -1,4 +1,4 @@
-/* $Id: devops.c,v 1.7 1997/03/18 17:58:19 jj Exp $
+/* $Id: devops.c,v 1.10 1997/05/14 20:44:59 davem Exp $
  * devops.c:  Device operations using the PROM.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -10,6 +10,9 @@
 #include <asm/openprom.h>
 #include <asm/oplib.h>
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
+
 /* Open the device described by the string 'dstr'.  Returns the handle
  * to that device used for subsequent operations on that device.
  * Returns -1 on failure.
@@ -35,7 +38,7 @@ prom_devopen(char *dstr)
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 
@@ -57,10 +60,11 @@ prom_devclose(int dhandle)
 		(*(romvec->pv_v2devops.v2_dev_close))(dhandle);
 		break;
         case PROM_AP1000:
+	default:
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	return 0;
@@ -83,10 +87,11 @@ prom_seek(int dhandle, unsigned int seekhi, unsigned int seeklo)
 		(*(romvec->pv_v2devops.v2_dev_seek))(dhandle, seekhi, seeklo);
 		break;
         case PROM_AP1000:
+	default:
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 
diff --git a/arch/sparc/prom/misc.c b/arch/sparc/prom/misc.c
index d23d93f61..fede033dd 100644
--- a/arch/sparc/prom/misc.c
+++ b/arch/sparc/prom/misc.c
@@ -1,4 +1,4 @@
-/* $Id: misc.c,v 1.13 1997/04/10 05:12:59 davem Exp $
+/* $Id: misc.c,v 1.15 1997/05/14 20:45:00 davem Exp $
  * misc.c:  Miscellaneous prom functions that don't belong
  *          anywhere else.
  *
@@ -13,6 +13,9 @@
 #include <asm/oplib.h>
 #include <asm/auxio.h>
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
+
 /* Reset and reboot the machine with the command 'bcommand'. */
 void
 prom_reboot(char *bcommand)
@@ -22,7 +25,7 @@ prom_reboot(char *bcommand)
 	(*(romvec->pv_reboot))(bcommand);
 	/* Never get here. */
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 }
@@ -40,7 +43,7 @@ prom_feval(char *fstring)
 	else
 		(*(romvec->pv_fortheval.v2_eval))(fstring);
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 }
@@ -72,7 +75,7 @@ prom_cmdline(void)
 	save_flags(flags); cli();
 	(*(romvec->pv_abort))();
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	install_linux_ticker();
@@ -97,7 +100,7 @@ again:
 	(*(romvec->pv_halt))();
 	/* Never get here. */
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	goto again; /* PROM is out to get me -DaveM */
diff --git a/arch/sparc/prom/mp.c b/arch/sparc/prom/mp.c
index cf71cca60..8f07f9d40 100644
--- a/arch/sparc/prom/mp.c
+++ b/arch/sparc/prom/mp.c
@@ -1,4 +1,4 @@
-/* $Id: mp.c,v 1.7 1997/03/18 17:58:23 jj Exp $
+/* $Id: mp.c,v 1.9 1997/05/14 20:45:01 davem Exp $
  * mp.c:  OpenBoot Prom Multiprocessor support routines.  Don't call
  *        these on a UP or else you will halt and catch fire. ;)
  *
@@ -12,6 +12,9 @@
 #include <asm/openprom.h>
 #include <asm/oplib.h>
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
+
 /* Start cpu with prom-tree node 'cpunode' using context described
  * by 'ctable_reg' in context 'ctx' at program counter 'pc'.
  *
@@ -36,7 +39,7 @@ prom_startcpu(int cpunode, struct linux_prom_registers *ctable_reg, int ctx, cha
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 
@@ -65,7 +68,7 @@ prom_stopcpu(int cpunode)
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 
@@ -94,7 +97,7 @@ prom_idlecpu(int cpunode)
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 
@@ -123,7 +126,7 @@ prom_restartcpu(int cpunode)
 		break;
 	};
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 
diff --git a/arch/sparc/prom/segment.c b/arch/sparc/prom/segment.c
index cb1def0e1..96b543727 100644
--- a/arch/sparc/prom/segment.c
+++ b/arch/sparc/prom/segment.c
@@ -1,4 +1,4 @@
-/* $Id: segment.c,v 1.3 1996/09/19 20:27:28 davem Exp $
+/* $Id: segment.c,v 1.5 1997/05/14 20:45:02 davem Exp $
  * segment.c:  Prom routine to map segments in other contexts before
  *             a standalone is completely mapped.  This is for sun4 and
  *             sun4c architectures only.
@@ -12,6 +12,9 @@
 #include <asm/openprom.h>
 #include <asm/oplib.h>
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
+
 /* Set physical segment 'segment' at virtual address 'vaddr' in
  * context 'ctx'.
  */
@@ -22,7 +25,7 @@ prom_putsegment(int ctx, unsigned long vaddr, int segment)
 	save_flags(flags); cli();
 	(*(romvec->pv_setctxt))(ctx, (char *) vaddr, segment);
 	__asm__ __volatile__("ld [%0], %%g6\n\t" : :
-			     "r" (&current_set[smp_processor_id()]) :
+			     "r" (&current_set[hard_smp_processor_id()]) :
 			     "memory");
 	restore_flags(flags);
 	return;
diff --git a/arch/sparc/prom/tree.c b/arch/sparc/prom/tree.c
index e5e2ceb3d..251b2ee6a 100644
--- a/arch/sparc/prom/tree.c
+++ b/arch/sparc/prom/tree.c
@@ -1,4 +1,4 @@
-/* $Id: tree.c,v 1.16 1997/03/19 14:53:16 davem Exp $
+/* $Id: tree.c,v 1.18 1997/05/14 20:45:03 davem Exp $
  * tree.c: Basic device tree traversal/scanning for the Linux
  *         prom library.
  *
@@ -15,10 +15,12 @@
 #include <asm/openprom.h>
 #include <asm/oplib.h>
 
+/* XXX Let's get rid of this thing if we can... */
+extern struct task_struct *current_set[NR_CPUS];
 
 /* Macro to restore "current" to the g6 register. */
 #define restore_current() __asm__ __volatile__("ld [%0], %%g6\n\t" : : \
-			  "r" (&current_set[smp_processor_id()]) : \
+			  "r" (&current_set[hard_smp_processor_id()]) : \
 			  "memory")
 
 static char promlib_buf[128];
diff --git a/arch/sparc64/Makefile b/arch/sparc64/Makefile
index bacf9b095..a70f9ebf8 100644
--- a/arch/sparc64/Makefile
+++ b/arch/sparc64/Makefile
@@ -1,4 +1,4 @@
-# $Id: Makefile,v 1.15 1997/04/14 17:04:49 jj Exp $
+# $Id: Makefile,v 1.16 1997/05/04 07:21:08 davem Exp $
 # sparc64/Makefile
 #
 # Makefile for the architecture dependent flags and dependencies on the
@@ -29,7 +29,7 @@ CFLAGS := $(CFLAGS) -pipe \
 
 LINKFLAGS = -T arch/sparc64/vmlinux.lds
 
-HEAD := arch/sparc64/kernel/head.o
+HEAD := arch/sparc64/kernel/head.o arch/sparc64/kernel/init_task.o
 
 SUBDIRS := $(SUBDIRS) arch/sparc64/kernel arch/sparc64/lib arch/sparc64/mm \
 	arch/sparc64/prom
diff --git a/arch/sparc64/defconfig b/arch/sparc64/defconfig
index 9ecbf90b4..fbc4a5073 100644
--- a/arch/sparc64/defconfig
+++ b/arch/sparc64/defconfig
@@ -40,6 +40,7 @@ SUN_FB_CGFOURTEEN=y
 SUN_FB_BWTWO=y
 SUN_FB_LEO=y
 TADPOLE_FB_WEITEK=y
+SUN_FB_CREATOR=y
 
 #
 # Misc Linux/SPARC drivers
@@ -132,8 +133,8 @@ CONFIG_SUNLANCE=y
 #
 # Filesystems
 #
-CONFIG_QUOTA=y
-CONFIG_MINIX_FS=y
+# CONFIG_QUOTA is not set
+# CONFIG_MINIX_FS is not set
 CONFIG_EXT2_FS=y
 # CONFIG_FAT_FS is not set
 # CONFIG_MSDOS_FS is not set
@@ -141,7 +142,9 @@ CONFIG_EXT2_FS=y
 # CONFIG_UMSDOS_FS is not set
 CONFIG_PROC_FS=y
 CONFIG_NFS_FS=y
-# CONFIG_ROOT_NFS is not set
+CONFIG_ROOT_NFS=y
+CONFIG_RNFS_BOOTP=y
+# CONFIG_RNFS_RARP is not set
 # CONFIG_NFSD is not set
 CONFIG_SUNRPC=y
 CONFIG_LOCKD=y
diff --git a/arch/sparc64/kernel/Makefile b/arch/sparc64/kernel/Makefile
index d66fa06e7..4a07295e1 100644
--- a/arch/sparc64/kernel/Makefile
+++ b/arch/sparc64/kernel/Makefile
@@ -1,4 +1,4 @@
-# $Id: Makefile,v 1.16 1997/04/17 20:35:37 jj Exp $
+# $Id: Makefile,v 1.20 1997/05/18 08:42:11 davem Exp $
 # Makefile for the linux kernel.
 #
 # Note! Dependencies are done automagically by 'make dep', which also
@@ -13,12 +13,12 @@
 .S.o:
 	$(CC) -D__ASSEMBLY__ -ansi -c $< -o $*.o
 
-all: kernel.o head.o
+all: kernel.o head.o init_task.o
 
 O_TARGET := kernel.o
 O_OBJS   := etrap.o rtrap.o hack.o process.o setup.o cpu.o idprom.o \
 	    systbls.o traps.o entry.o devices.o auxio.o ioport.o \
-	    irq.o time.o sys_sparc.o
+	    irq.o time.o sys_sparc.o signal.o winfixup.o
 OX_OBJS  := sparc64_ksyms.o
 
 ifdef CONFIG_SPARC32_COMPAT
@@ -26,12 +26,17 @@ ifdef CONFIG_SPARC32_COMPAT
 endif
 
 ifdef CONFIG_BINFMT_ELF32
-  O_OBJS += sparcelf32.o
+  O_OBJS += binfmt_elf32.o
 endif
 
 head.o: head.S ttable.S itlb_miss.S dtlb_miss.S dtlb_prot.S
 	$(CC) -D__ASSEMBLY__ -ansi -c $*.S -o $*.o
 
+#
+# This is just to get the dependencies...
+#
+binfmt_elf32.o: $(TOPDIR)/fs/binfmt_elf.c
+
 check_asm: dummy
 	@echo "#include <linux/sched.h>" > tmp.c
 	$(CC) -E tmp.c -o tmp.i
diff --git a/arch/sparc64/kernel/binfmt_elf32.c b/arch/sparc64/kernel/binfmt_elf32.c
new file mode 100644
index 000000000..05d50fe56
--- /dev/null
+++ b/arch/sparc64/kernel/binfmt_elf32.c
@@ -0,0 +1,34 @@
+/* binfmt_elf32.c: Support 32-bit Sparc ELF binaries on Ultra.
+ *
+ */
+
+#define ELF_ARCH		EM_SPARC
+#define ELF_CLASS		ELFCLASS32
+#define ELF_DATA		ELFDATA2MSB;
+
+#include <asm/processor.h>
+#include <linux/module.h>
+#include <linux/config.h>
+
+#define elf_addr_t	u32
+#define elf_caddr_t	u32
+#undef start_thread
+#define start_thread start_thread32
+#define init_elf_binfmt init_elf32_binfmt
+#undef CONFIG_BINFMT_ELF
+#ifdef CONFIG_BINFMT_ELF32
+#define CONFIG_BINFMT_ELF CONFIG_BINFMT_ELF32
+#endif
+#undef CONFIG_BINFMT_ELF_MODULE
+#ifdef CONFIG_BINFMT_ELF32_MODULE
+#define CONFIG_BINFMT_ELF_MODULE CONFIG_BINFMT_ELF32_MODULE
+#endif
+#define ELF_FLAGS_INIT	current->tss.flags |= SPARC_FLAG_32BIT
+
+MODULE_DESCRIPTION("Binary format loader for compatibility with 32bit SparcLinux binaries on the Ultra");
+MODULE_AUTHOR("Eric Youngdale, David S. Miller, Jakub Jelinek");
+
+#undef MODULE_DESCRIPTION
+#undef MODULE_AUTHOR
+
+#include "../../../fs/binfmt_elf.c"
diff --git a/arch/sparc64/kernel/dtlb_prot.S b/arch/sparc64/kernel/dtlb_prot.S
index 8eec19260..2cf372ca9 100644
--- a/arch/sparc64/kernel/dtlb_prot.S
+++ b/arch/sparc64/kernel/dtlb_prot.S
@@ -1,4 +1,4 @@
-/* $Id: dtlb_prot.S,v 1.10 1997/03/25 09:47:13 davem Exp $
+/* $Id: dtlb_prot.S,v 1.12 1997/05/18 10:04:43 davem Exp $
  * dtlb_prot.S:	Data TLB protection code, this is included directly
  *              into the trap table.
  *
@@ -18,30 +18,30 @@
   /*0x04*/	srlx		%g1, 8, %g3			! Position PGD offset
   /*0x08*/	sllx		%g1, 2, %g4			! Position PMD offset
   /*0x0c*/	and		%g3, %g2, %g3			! Mask PGD offset
-  /*0x10*/	and		%g4, %g2, %g3			! Mask PMD offset
+  /*0x10*/	and		%g4, %g2, %g4			! Mask PMD offset
   /*0x14*/	ldxa		[%g7 + %g3] ASI_PHYS_USE_EC, %g5	! Load PGD
-  /*0x18*/	ldxa		[%g5 + %g3] ASI_PHYS_USE_EC, %g4	! Load PMD
+  /*0x18*/	ldxa		[%g5 + %g4] ASI_PHYS_USE_EC, %g4	! Load PMD
   /*0x1c*/	ldxa		[%g0] ASI_DMMU_TSB_8KB_PTR, %g1		! For PTE offset
 
 				/* ICACHE line 2 */
   /*0x20*/	srlx		%g1, 1, %g1				! PTE offset
   /*0x24*/	ldxa		[%g4 + %g1] ASI_PHYS_USE_EC, %g3	! Load PTE
   /*0x28*/	andcc		%g3, _PAGE_WRITE, %g0			! Writable?
-  /*0x2c*/	be,pt		%xcc, sparc64_dtlb_fault		! Nope...
+  /*0x2c*/	be,pt		%xcc, sparc64_dtlb_prot_catch		! Nope...
   /*0x30*/	 or		%g3, (MODIFIED_BITS), %g3		! Yes it is
   /*0x34*/	mov		TLB_TAG_ACCESS, %g5			! Get the page
-  /*0x38*/	ldxa		[%g5] ASI_DMMU, %g1			! From MMU
-  /*0x3c*/	add		%g2, 7, %g5				! Compute mask
+  /*0x38*/	add		%g1, %g4, %g1				! to get a tmpreg
+  /*0x3c*/	ldxa		[%g5] ASI_DMMU, %g4			! From MMU
 
 				/* ICACHE line 3 */
-  /*0x40*/	andn		%g1, %g5, %g1				! Mask page
-  /*0x44*/	or		%g1, 0x10, %g1				! 2ndary Context
-  /*0x48*/	stxa		%g0, [%g1] ASI_DMMU_DEMAP		! TLB flush page
-  /*0x4c*/	membar		#Sync					! Synchronize
-  /*0x50*/	stxa		%g3, [%g4 + %g1] ASI_PHYS_USE_EC	! Update sw PTE
-  /*0x54*/	stxa		%g3, [%g0] ASI_DTLB_DATA_IN		! TLB load
-  /*0x58*/	retry							! Trap return
-  /*0x5c*/	nop
+  /*0x40*/	add		%g2, 7, %g5				! Compute mask
+  /*0x44*/	andn		%g4, %g5, %g4				! Mask page
+  /*0x48*/	or		%g4, 0x10, %g4				! 2ndary Context
+  /*0x4c*/	stxa		%g0, [%g4] ASI_DMMU_DEMAP		! TLB flush page
+  /*0x50*/	membar		#Sync					! Synchronize
+  /*0x54*/	stxa		%g3, [%g1] ASI_PHYS_USE_EC		! Update sw PTE
+  /*0x58*/	stxa		%g3, [%g0] ASI_DTLB_DATA_IN		! TLB load
+  /*0x5c*/	retry							! Trap return
 
 				/* ICACHE line 4 */
   /*0x60*/	nop
diff --git a/arch/sparc64/kernel/entry.S b/arch/sparc64/kernel/entry.S
index 16fe5c8a0..589d1661a 100644
--- a/arch/sparc64/kernel/entry.S
+++ b/arch/sparc64/kernel/entry.S
@@ -1,4 +1,4 @@
-/* $Id: entry.S,v 1.14 1997/04/14 06:56:54 davem Exp $
+/* $Id: entry.S,v 1.21 1997/05/18 10:04:44 davem Exp $
  * arch/sparc64/kernel/entry.S:  Sparc64 trap low-level entry points.
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -24,31 +24,70 @@
 
 	.text
         .align  4
-/* FIXME: This is still debugging hack */
-	.globl	sparc64_dtlb_fault, sparc64_dtlb_refbit_catch, sparc64_itlb_refbit_catch
-sparc64_dtlb_fault:
+	.globl	sparc64_dtlb_prot_catch, sparc64_dtlb_refbit_catch
+	.globl	sparc64_itlb_refbit_catch
+
+	/* Note, DMMU SFAR not updated for fast tlb data access miss
+	 * traps, so we must use tag access to find the right page.
+	 * However for DMMU fast protection traps it is updated so
+	 * we use, but we must also clear it _before_ we enable interrupts
+	 * and save state because there is a race where we can push a user
+	 * window right now in etrap, a protection fault happens (for example
+	 * to update the dirty bit) and since we left crap in the sfsr
+	 * it will not get updated properly.
+	 */
+sparc64_dtlb_prot_catch:
+	wr	%g0, ASI_DMMU, %asi
 	rdpr	%pstate, %g1
 	wrpr	%g1, PSTATE_AG|PSTATE_MG, %pstate
+	rdpr	%tl, %g2
+	ldxa	[%g0 + TLB_TAG_ACCESS] %asi, %g5
+	ldxa	[%g0 + TLB_SFSR] %asi, %g4
+	cmp	%g2, 1
+	stxa	%g0, [%g0 + TLB_SFSR] %asi
+	bgu,a	%icc, winfix_trampoline
+	 rdpr	%tpc, %g5
 	ba,pt	%xcc, etrap
 	 rd	%pc, %g7
-	call	sparc64_dtlb_fault_handler
-	 nop
+	b,a,pt	%xcc, 1f
 
 sparc64_dtlb_refbit_catch:
+	wr	%g0, ASI_DMMU, %asi
 	rdpr	%pstate, %g1
 	wrpr	%g1, PSTATE_AG|PSTATE_MG, %pstate
+	rdpr	%tl, %g2
+	ldxa	[%g0 + TLB_TAG_ACCESS] %asi, %g5
+	cmp	%g2, 1
+	clr	%g4				! sfsr not updated for tlb misses
+	bgu,a	%icc, winfix_trampoline
+	 rdpr	%tpc, %g5
 	ba,pt	%xcc, etrap
 	 rd	%pc, %g7
-	call	sparc64_dtlb_refbit_handler
-	 add	%sp, STACK_BIAS + REGWIN_SZ, %o0
+1:
+	mov	%l5, %o4				! raw tag access
+	mov	%l4, %o5				! raw sfsr
+	srlx	%l5, PAGE_SHIFT, %o3
+	clr	%o1					! text_fault == 0
+	sllx	%o3, PAGE_SHIFT, %o3			! address
+	and	%l4, 0x4, %o2				! write == sfsr.W
+	call	do_sparc64_fault
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o0	! pt_regs ptr
+	ba,a,pt	%xcc, rtrap
 
 sparc64_itlb_refbit_catch:
 	rdpr	%pstate, %g1
 	wrpr	%g1, PSTATE_AG|PSTATE_MG, %pstate
 	ba,pt	%xcc, etrap
 	 rd	%pc, %g7
-	call	sparc64_dtlb_refbit_handler
-	 nop
+
+	ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TPC], %o3
+	mov	1, %o1					! text_fault == 1
+	clr	%o2					! write == 0
+	clr	%o4					! tag access (N/A)
+	clr	%o5					! raw sfsr (N/A)
+	call	do_sparc64_fault
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o0	! pt_regs ptr
+	ba,a,pt	%xcc, rtrap
 
 	/* Note check out head.h, this code isn't even used for UP,
 	 * for SMP things will be different.  In particular the data
@@ -121,6 +160,7 @@ breakpoint_trap:
 	ba,a,pt	%xcc, rtrap
 
 	.globl	sys_pipe, sys_execve, sys_sigpause, sys_nis_syscall
+	.globl	sys_sigsuspend, sys_sigreturn
 
 sys_pipe:
 	sethi	%hi(sparc_pipe), %g1
@@ -147,12 +187,35 @@ sys_sigpause:
 	
 	ld	[%curptr + AOFF_task_flags], %l5
 	andcc	%l5, 0x20, %g0
+	be,pt	%icc, rtrap
+	 nop
+	call	syscall_trace
+	 nop
+	ba,a,pt	%xcc, rtrap
+
+sys_sigsuspend:
+	call	do_sigsuspend
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o0
+	
+	ld	[%curptr + AOFF_task_flags], %l5
+	andcc	%l5, 0x20, %g0
 	be,pt	%icc, ret_sys_call
-	 clr	%o0
+	 nop
 	call	syscall_trace
 	 nop
-	ba,pt	%xcc, ret_sys_call
-	 clr	%o0
+	ba,a,pt	%xcc, rtrap
+	
+sys_sigreturn:
+	call	do_sigreturn
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o0
+	
+	ld	[%curptr + AOFF_task_flags], %l5
+	andcc	%l5, 0x20, %g0
+	be,pt	%icc, ret_sys_call
+	 nop
+	call	syscall_trace
+	 nop
+	ba,a,pt	%xcc, rtrap
 
 	/* This is how fork() was meant to be done, 11 instruction entry. -DaveM */
 	.globl	sys_fork, sys_vfork, sys_clone
@@ -200,7 +263,7 @@ linux_syscall_trace:
 	.globl	ret_from_syscall
 ret_from_syscall:
 	ba,pt	%xcc, ret_sys_call
-	 ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_I0], %o0
+	 ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_V9_I0], %o0
 
 	/* Linux native and SunOS system calls enter here... */
 	.align	4
@@ -232,13 +295,18 @@ syscall_is_too_hard:
 	call	%l7
 	 mov	%i5, %o5
 
-	stx	%o0, [%sp + STACK_BIAS + REGWIN_SZ + PT_I0]
-
+	stx	%o0, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_I0]
+#if 0
+	/* Debugging... */
+	call	syscall_trace_exit
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o0
+	ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_V9_I0], %o0
+#endif
 	.globl	ret_sys_call
 ret_sys_call:
 	ldx	[%curptr + AOFF_task_flags], %l6
 	mov	%ulo(TSTATE_XCARRY | TSTATE_ICARRY), %g2
-	ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_TSTATE], %g3
+	ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TSTATE], %g3
 	cmp	%o0, -ENOIOCTLCMD
 	sllx	%g2, 32, %g2
 	bgeu,pn	%xcc, 1f
@@ -247,34 +315,34 @@ ret_sys_call:
 	/* System call success, clear Carry condition code. */
 	andn	%g3, %g2, %g3
 	clr	%l6
-	stx	%g3, [%sp + STACK_BIAS + REGWIN_SZ + PT_TSTATE]	
+	stx	%g3, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TSTATE]	
 	bne,pn	%icc, linux_syscall_trace2
-	 ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_TNPC], %l1 /* pc = npc */
+	 ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TNPC], %l1 /* pc = npc */
 	add	%l1, 0x4, %l2				      /* npc = npc+4 */
-	stx	%l1, [%sp + STACK_BIAS + REGWIN_SZ + PT_TPC]
+	stx	%l1, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TPC]
 	ba,pt	%xcc, rtrap
-	 stx	%l2, [%sp + STACK_BIAS + REGWIN_SZ + PT_TNPC]
+	 stx	%l2, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TNPC]
 1:
 	/* System call failure, set Carry condition code.
 	 * Also, get abs(errno) to return to the process.
 	 */
 	sub	%g0, %o0, %o0
 	or	%g3, %g2, %g3
-	stx	%o0, [%sp + STACK_BIAS + REGWIN_SZ + PT_I0]
+	stx	%o0, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_I0]
 	mov	1, %l6
-	stx	%g3, [%sp + STACK_BIAS + REGWIN_SZ + PT_TSTATE]
+	stx	%g3, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TSTATE]
 	bne,pn	%icc, linux_syscall_trace2
-	 ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_TNPC], %l1 /* pc = npc */
+	 ldx	[%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TNPC], %l1 /* pc = npc */
 	add	%l1, 0x4, %l2				      /* npc = npc+4 */
-	stx	%l1, [%sp + STACK_BIAS + REGWIN_SZ + PT_TPC]
+	stx	%l1, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TPC]
 	ba,pt	%xcc, rtrap
-	 stx	%l2, [%sp + STACK_BIAS + REGWIN_SZ + PT_TNPC]
+	 stx	%l2, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TNPC]
 
 linux_syscall_trace2:
 	call	syscall_trace
 	 add	%l1, 0x4, %l2			/* npc = npc+4 */
-	stx	%l1, [%sp + STACK_BIAS + REGWIN_SZ + PT_TPC]
+	stx	%l1, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TPC]
 	ba,pt	%xcc, rtrap
-	 stx	%l2, [%sp + STACK_BIAS + REGWIN_SZ + PT_TNPC]
+	 stx	%l2, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_TNPC]
 
 /* End of entry.S */
diff --git a/arch/sparc64/kernel/etrap.S b/arch/sparc64/kernel/etrap.S
index 7d293a88b..f936b3071 100644
--- a/arch/sparc64/kernel/etrap.S
+++ b/arch/sparc64/kernel/etrap.S
@@ -1,34 +1,39 @@
-/* $Id: etrap.S,v 1.11 1997/04/14 17:04:45 jj Exp $
+/* $Id: etrap.S,v 1.17 1997/05/18 22:52:09 davem Exp $
  * etrap.S: Preparing for entry into the kernel on Sparc V9.
  *
- * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
+ * Copyright (C) 1996, 1997 David S. Miller (davem@caip.rutgers.edu)
  * Copyright (C) 1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
  */
 
 #include <asm/asi.h>
 #include <asm/pstate.h>
 #include <asm/ptrace.h>
+#include <asm/page.h>
 #include <asm/spitfire.h>
 #include <asm/head.h>
 
-/* We assume that pstate, when entering this, has AG and IE bits set, MG and IG clear */
+	/* We assume that pstate, when entering this, has AG and
+	 * IE bits set, MG and IG clear.
+	 *
+	 * We also guarentee for caller that AG %g4 and %g5 will have
+	 * their values preserved and left in %l4 and %l5 respectively
+	 * for him (fault handling needs this).
+	 */
 
 	.text
 	.align	32
 	.globl	etrap, etrap_irq
 etrap:
-	rdpr		%pil, %g4
+	rdpr		%pil, %g2
 etrap_irq:
 	rdpr		%tstate, %g1
-	sllx		%g4, 20, %g4
-	rdpr		%tpc, %g2
-	or		%g1, %g4, %g1
-	rdpr		%tnpc, %g3
+	sllx		%g2, 20, %g2
+	or		%g1, %g2, %g1
 
 	/* What happens more often? etrap when already in priv or from userland? */
 	andcc		%g1, TSTATE_PRIV, %g0
 	bne,a,pn	%xcc, 1f
-	 sub		%sp, REGWIN_SZ + TRACEREG_SZ - STACK_BIAS, %g5
+	 sub		%sp, REGWIN_SZ + TRACEREG_SZ - STACK_BIAS, %g2
 
 	/* Just when going from userland to privileged mode,
 	 * we have to change this stuff.
@@ -38,41 +43,48 @@ etrap_irq:
 	 * trap level until PRIMARY_CONTEXT is set to zero, else
 	 * we fall out of NUCLEUS too soon and crash hard.
 	 */
-	rdpr		%wstate, %g5
-	mov		PRIMARY_CONTEXT, %g7
-	ldxa		[%g7] ASI_DMMU, %g4
-	mov		SECONDARY_CONTEXT, %g6
-	stxa		%g0, [%g7] ASI_DMMU
-	stxa		%g4, [%g6] ASI_DMMU
-	wrpr		%g0, 0x0, %tl
+	mov		PRIMARY_CONTEXT, %g1
+	ldxa		[%g1] ASI_DMMU, %g2
+	stxa		%g0, [%g1] ASI_DMMU
 
-	sll		%g5, 3, %g5
-	sethi		%uhi(KERNBASE), %g4
-	or		%g4, %ulo(KERNBASE), %g4
-	sethi		%hi(current_set), %g6
-	or		%g6, %lo(current_set), %g6
-	sllx		%g4, 32, %g4
-	wrpr		%g5, %wstate
-	rdpr		%canrestore, %g5
-	ldx		[%g6 + %g4], %g6
-#ifdef __SMP__
-/* FIXME: Fix the above insn for SMP */
-#endif
-	wrpr		%g0, 0, %canrestore
-	wrpr		%g5, 0, %otherwin
-	ba,pt		%xcc, 2f
-	 ldx		[%g6 + AOFF_task_saved_kernel_stack], %g5
+	mov		SECONDARY_CONTEXT, %g1
+	stxa		%g2, [%g1] ASI_DMMU
+
+	rd		%pic, %g1
+	sethi		%hi((PAGE_SIZE<<1)-TRACEREG_SZ-REGWIN_SZ), %g2
+	or		%g2, %lo((PAGE_SIZE<<1)-TRACEREG_SZ-REGWIN_SZ), %g2
+	add		%g1, %g2, %g2
+	rdpr		%tstate, %g1
 1:
+	stx		%g1, [%g2 + REGWIN_SZ + PT_V9_TSTATE]
+	rdpr		%tpc, %g1
+	rdpr		%tnpc, %g3
+	stx		%g1, [%g2 + REGWIN_SZ + PT_V9_TPC]
+	rd		%y, %g1
+	stx		%g3, [%g2 + REGWIN_SZ + PT_V9_TNPC]
+	stx		%g1, [%g2 + REGWIN_SZ + PT_V9_Y]
+
 	wrpr		%g0, 0x0, %tl
-2:
-	rd		%y, %g4
-	stx		%g1, [%g5 + REGWIN_SZ + PT_V9_TSTATE]
-	stx		%g2, [%g5 + REGWIN_SZ + PT_V9_TPC]
-	stx		%g3, [%g5 + REGWIN_SZ + PT_V9_TNPC]
-	stx		%g4, [%g5 + REGWIN_SZ + PT_V9_Y]
 	rdpr		%pstate, %g1
-	save		%g5, -STACK_BIAS, %sp
+	save		%g2, -STACK_BIAS, %sp
+
+	/* Must guarentee that here andcc of TSTATE_PRIV at the top is
+	 * still valid in %ccr register.  Don't show this trick to your
+	 * mom. -DaveM
+	 */
+	bne,pn		%xcc, 1f
+	 rdpr		%canrestore, %g3
+	wrpr		%g0, 0, %canrestore
+	wrpr		%g3, 0, %otherwin
+
+	rdpr		%wstate, %g6
+	sll		%g6, 3, %g6
+	wrpr		%g6, %wstate
+
+1:
 	mov		%g1, %l1
+	mov		%g4, %l4
+	mov		%g5, %l5
 	mov		%g7, %l2
 	wrpr		%l1, PSTATE_AG, %pstate
 	stx		%g1, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_G1]
@@ -91,24 +103,13 @@ etrap_irq:
 	stx		%i6, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_I6]
 	stx		%i7, [%sp + STACK_BIAS + REGWIN_SZ + PT_V9_I7]
 	wrpr		%l1, (PSTATE_IE | PSTATE_AG), %pstate
-	sethi		%uhi(KERNBASE), %g4
-	or		%g4, %ulo(KERNBASE), %g4
-	sethi		%hi(current_set), %g6
-	or		%g6, %lo(current_set), %g6
-	sllx		%g4, 32, %g4
+	srlx		%sp, 43, %g4
+	rd		%pic, %g6
 	jmpl		%l2 + 0x4, %g0
-	 ldx		[%g6 + %g4], %g6
-#ifdef __SMP__
-/* FIXME: Fix the above insn for SMP */
-#endif
+	 sllx		%g4, 43, %g4
 
 	.globl	etraptl1
 etraptl1:
-	rdpr	%tl, %g4
 	rdpr	%tstate, %g1
-	sub	%g4, 1, %g4
-	rdpr	%tpc, %g2
-	rdpr	%tnpc, %g3
-	wrpr	%g4, 0x0, %tl
 	ba,pt	%xcc, 1b
-	 sub	%sp, REGWIN_SZ + TRACEREG_SZ - STACK_BIAS, %g5
+	 sub	%sp, REGWIN_SZ + TRACEREG_SZ - STACK_BIAS, %g2
diff --git a/arch/sparc64/kernel/hack.S b/arch/sparc64/kernel/hack.S
index 0aca22a77..034bda2d0 100644
--- a/arch/sparc64/kernel/hack.S
+++ b/arch/sparc64/kernel/hack.S
@@ -4,8 +4,6 @@
    to compile... */
 	.text
 	.align	8
-	.globl	_sigpause_common
-_sigpause_common:	retl;nop
 	.globl	breakpoint
 breakpoint:	retl;nop
 	.globl	do_cee
@@ -36,8 +34,6 @@ do_fpother_tl1:	retl;nop
 do_iae:	retl;nop
 	.globl	do_iae_tl1
 do_iae_tl1:	retl;nop
-	.globl	do_ill
-do_ill:	retl;nop
 	.globl	do_ill_tl1
 do_ill_tl1:	retl;nop
 	.globl	do_irq
@@ -48,8 +44,6 @@ do_irq_tl1:	retl;nop
 do_lddfmna:	retl;nop
 	.globl	do_lddfmna_tl1
 do_lddfmna_tl1:	retl;nop
-	.globl	do_mna
-do_mna:	retl;nop
 	.globl	do_mna_tl1
 do_mna_tl1:	retl;nop
 	.globl	do_paw
@@ -60,8 +54,6 @@ do_paw_tl1:	retl;nop
 do_privact:	retl;nop
 	.globl	do_privop
 do_privop:	retl;nop
-	.globl	do_signal
-do_signal:	retl;nop
 	.globl	do_stdfmna
 do_stdfmna:	retl;nop
 	.globl	do_stdfmna_tl1
@@ -200,15 +192,7 @@ sunos_write:	retl;nop
 sunos_writev:	retl;nop
 	.globl	sys_ptrace
 sys_ptrace:	retl;nop
-	.globl	sys_sigreturn
-sys_sigreturn:	retl;nop
-	.globl	sys_sigstack
-sys_sigstack:	retl;nop
-	.globl	sys_sigsuspend
-sys_sigsuspend:	retl;nop
 	.globl	syscall_trace
 syscall_trace:	retl;nop
 	.globl	sys32_ptrace
 sys32_ptrace:	retl;nop
-	.globl	do_sigpause
-do_sigpause:	retl;nop
diff --git a/arch/sparc64/kernel/head.S b/arch/sparc64/kernel/head.S
index fdbe87aa3..4babe3eb4 100644
--- a/arch/sparc64/kernel/head.S
+++ b/arch/sparc64/kernel/head.S
@@ -1,4 +1,4 @@
-/* $Id: head.S,v 1.27 1997/04/04 00:49:49 davem Exp $
+/* $Id: head.S,v 1.30 1997/05/18 22:52:12 davem Exp $
  * head.S: Initial boot code for the Sparc64 port of Linux.
  *
  * Copyright (C) 1996,1997 David S. Miller (davem@caip.rutgers.edu)
@@ -262,16 +262,22 @@ sun4u_init:
 	stx     %g6, [%g2 + %g4]
 	stx     %g5, [%g3 + %g4]
 
-	sethi	%hi(init_task), %g6
-	or	%g6, %lo(init_task), %g6
+	sethi	%hi(init_task_union), %g6
+	or	%g6, %lo(init_task_union), %g6
 	add	%g6, %g4, %g6			! g6 usage is fixed as well
 	mov	%sp, %l6
 	mov	%o4, %l7
 
-	sethi	%hi(bootup_kernel_stack + 0x2000 - STACK_BIAS - REGWIN_SZ), %g5
-	or	%g5, %lo(bootup_kernel_stack + 0x2000 - STACK_BIAS - REGWIN_SZ), %g5
-	add	%g5, %g4, %sp
+	/* Setup "Linux Current Register", thanks Sun 8-) */
+	wr	%g0, 0x1, %pcr
+	wr	%g6, 0x0, %pic
+
+	mov	1, %g5
+	sllx	%g5, (PAGE_SHIFT + 1), %g5
+	sub	%g5, (REGWIN_SZ + STACK_BIAS), %g5
+	add	%g6, %g5, %sp
 	mov	0, %fp
+
 	wrpr	%g0, 0, %wstate
 	wrpr	%g0, 0x0, %tl
 
@@ -287,14 +293,20 @@ sun4u_init:
 	add	%l2, 1, %l2
 	add	%l0, %g4, %o0
 1:
-	call	bzero_1page
+	clr	%o1
+	sethi	%hi(PAGE_SIZE), %o2
+	or	%o2, %lo(PAGE_SIZE), %o2
+	call	__memset
 	 add	%l0, %l2, %l0
 	cmp	%l0, %l1
 	blu,pt	%xcc, 1b
 	 add	%l0, %g4, %o0
 
 	/* Now clear empty_zero_page */
-	call	bzero_1page
+	clr	%o1
+	sethi	%hi(PAGE_SIZE), %o2
+	or	%o2, %lo(PAGE_SIZE), %o2
+	call	__memset
 	 mov	%g4, %o0
 
 	mov	%l6, %o1			! OpenPROM stack
@@ -361,6 +373,9 @@ bootup_kernel_stack:
 
 #include "ttable.S"
 
+	/* This is just anal retentiveness on my part... */
+	.align	16384
+
 	.data
 	.align	8
 	.globl	nwindows, nwindowsm1 
diff --git a/arch/sparc64/kernel/init_task.c b/arch/sparc64/kernel/init_task.c
new file mode 100644
index 000000000..d0fc09346
--- /dev/null
+++ b/arch/sparc64/kernel/init_task.c
@@ -0,0 +1,18 @@
+#include <linux/mm.h>
+#include <linux/sched.h>
+
+#include <asm/pgtable.h>
+#include <asm/uaccess.h>
+
+static struct vm_area_struct init_mmap = INIT_MMAP;
+static struct fs_struct init_fs = INIT_FS;
+static struct files_struct init_files = INIT_FILES;
+static struct signal_struct init_signals = INIT_SIGNALS;
+struct mm_struct init_mm = INIT_MM;
+
+/* .text section in head.S is aligned at 2 page boundry and this gets linked
+ * right after that so that the init_task_union is aligned properly as well.
+ * We really don't need this special alignment like the Intel does, but
+ * I do it anyways for completeness.
+ */
+union task_union init_task_union __attribute__((__section__(".text"))) = { INIT_TASK };
diff --git a/arch/sparc64/kernel/process.c b/arch/sparc64/kernel/process.c
index 5ec62a6b6..4e2bbe016 100644
--- a/arch/sparc64/kernel/process.c
+++ b/arch/sparc64/kernel/process.c
@@ -1,4 +1,4 @@
-/*  $Id: process.c,v 1.6 1997/04/07 18:57:07 jj Exp $
+/*  $Id: process.c,v 1.11 1997/05/18 22:52:19 davem Exp $
  *  arch/sparc64/kernel/process.c
  *
  *  Copyright (C) 1995, 1996 David S. Miller (davem@caip.rutgers.edu)
@@ -346,7 +346,14 @@ void flush_thread(void)
 	}
 	
 	/* Now, this task is no longer a kernel thread. */
-	current->tss.flags &= ~SPARC_FLAG_KTHREAD;
+	if(current->tss.flags & SPARC_FLAG_KTHREAD) {
+		current->tss.flags &= ~SPARC_FLAG_KTHREAD;
+
+		/* exec_mmap() set context to NO_CONTEXT, here is
+		 * where we grab a new one.
+		 */
+		get_mmu_context(current);
+	}
 	current->tss.current_ds = USER_DS;
 }
 
@@ -416,6 +423,64 @@ clone_stackframe(struct sparc_stackf *dst, struct sparc_stackf *src)
 	return sp;
 }
 
+/* Standard stuff. */
+static inline void shift_window_buffer(int first_win, int last_win,
+				       struct thread_struct *tp)
+{
+	int i;
+
+	for(i = first_win; i < last_win; i++) {
+		tp->rwbuf_stkptrs[i] = tp->rwbuf_stkptrs[i+1];
+		memcpy(&tp->reg_window[i], &tp->reg_window[i+1],
+		       sizeof(struct reg_window));
+	}
+}
+
+void synchronize_user_stack(void)
+{
+	struct thread_struct *tp = &current->tss;
+	unsigned long window = tp->w_saved;
+
+	flush_user_windows();
+	if(window) {
+		int winsize = REGWIN_SZ;
+
+		if(tp->flags & SPARC_FLAG_32BIT)
+			winsize = REGWIN32_SZ;
+
+		window -= 1;
+		do {
+			unsigned long sp = tp->rwbuf_stkptrs[window];
+			struct reg_window *rwin = &tp->reg_window[window];
+
+			if(!copy_to_user((char *)sp, rwin, winsize)) {
+				shift_window_buffer(window, tp->w_saved - 1, tp);
+				tp->w_saved--;
+			}
+		} while(window--);
+	}
+}
+
+void fault_in_user_windows(struct pt_regs *regs)
+{
+	struct thread_struct *tp = &current->tss;
+	unsigned long window = tp->w_saved;
+	int winsize = REGWIN_SZ;
+
+	if(tp->flags & SPARC_FLAG_32BIT)
+		winsize = REGWIN32_SZ;
+	if(window) {
+		window -= 1;
+		do {
+			unsigned long sp = tp->rwbuf_stkptrs[window];
+			struct reg_window *rwin = &tp->reg_window[window];
+
+			if(copy_to_user((char *)sp, rwin, winsize))
+				do_exit(SIGILL);
+		} while(window--);
+	}
+	current->tss.w_saved = 0;
+}
 
 /* Copy a Sparc thread.  The fork() return value conventions
  * under SunOS are nothing short of bletcherous:
@@ -453,19 +518,18 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 #endif	
 
 	/* Calculate offset to stack_frame & pt_regs */
-	stack_offset = (PAGE_SIZE - TRACEREG_SZ);
+	stack_offset = ((PAGE_SIZE<<1) - TRACEREG_SZ);
 
 	if(regs->tstate & TSTATE_PRIV)
 		stack_offset -= REGWIN_SZ;
 
-	childregs = ((struct pt_regs *) (p->kernel_stack_page + stack_offset));
+	childregs = ((struct pt_regs *) (((unsigned long)p) + stack_offset));
 	*childregs = *regs;
 	new_stack = (((struct reg_window *) childregs) - 1);
 	old_stack = (((struct reg_window *) regs) - 1);
 	*new_stack = *old_stack;
 
-	p->saved_kernel_stack = ((unsigned long) new_stack);
-	p->tss.ksp = p->saved_kernel_stack - STACK_BIAS;
+	p->tss.ksp = ((unsigned long) new_stack) - STACK_BIAS;
 	p->tss.kpc = ((unsigned long) ret_from_syscall) - 0x8;
 	p->tss.kregs = childregs;
 
@@ -485,7 +549,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 		p->tss.current_ds = USER_DS;
 
 #if 0
-		if (sp != current->tss.kregs->u_regs[UREG_FP]) {
+		if (sp != regs->u_regs[UREG_FP]) {
 			struct sparc_stackf *childstack;
 			struct sparc_stackf *parentstack;
 
@@ -494,8 +558,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long sp,
 			 * Set some valid stack frames to give to the child.
 			 */
 			childstack = (struct sparc_stackf *)sp;
-			parentstack = (struct sparc_stackf *)
-					current->tss.kregs->u_regs[UREG_FP];
+			parentstack = (struct sparc_stackf *)regs->u_regs[UREG_FP];
 
 #if 0
 			printk("clone: parent stack:\n");
diff --git a/arch/sparc64/kernel/rtrap.S b/arch/sparc64/kernel/rtrap.S
index dfebd7ad8..1c4c4df04 100644
--- a/arch/sparc64/kernel/rtrap.S
+++ b/arch/sparc64/kernel/rtrap.S
@@ -1,4 +1,4 @@
-/* $Id: rtrap.S,v 1.11 1997/04/03 13:03:50 davem Exp $
+/* $Id: rtrap.S,v 1.14 1997/05/18 08:42:14 davem Exp $
  * rtrap.S: Preparing for return from trap on Sparc V9.
  *
  * Copyright (C) 1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
@@ -71,12 +71,11 @@ rtrap:
 	/* We came here from to_user, ie. we have now AG.
 	 * Also have to push user context back into primary.
 	 */
-	restore
-
 	mov	SECONDARY_CONTEXT, %g6
 	mov	PRIMARY_CONTEXT, %g7
 	ldxa	[%g6] ASI_DMMU, %g4
 	stxa	%g4, [%g7] ASI_DMMU
+	membar	#Sync			/* XXX flushi would be better -DaveM */
 
 	rdpr	%wstate, %g1
 	rdpr	%otherwin, %g2
@@ -84,7 +83,6 @@ rtrap:
 	wrpr	%g2, %g0, %canrestore
 	wrpr	%g1, %g0, %wstate
 	wrpr	%g0, %g0, %otherwin
-	retry
 1:
 	restore
 	retry
@@ -101,9 +99,14 @@ to_user:
 	ldx	[%g6 + AOFF_task_blocked], %o0
 	or	%l7, PSTATE_AG, %l7		! Will need this for setting back wstate
 	andncc	%l0, %o0, %g0
-	be,pt	%xcc, 3b
+	be,pt	%xcc, check_user_wins
 	 mov	%l5, %o2  
 	mov	%l6, %o3
-	add	%sp, STACK_BIAS + REGWIN_SZ, %o1
 	call	do_signal
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o1
+check_user_wins:
+	ldx	[%g6 + AOFF_task_tss + AOFF_thread_w_saved], %o2
+	brz,pt	%o2, 3b
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o1
+	call	fault_in_user_windows
 	 add	%o7, 3b-.-4, %o7
diff --git a/arch/sparc64/kernel/setup.c b/arch/sparc64/kernel/setup.c
index 170e5563e..47d900977 100644
--- a/arch/sparc64/kernel/setup.c
+++ b/arch/sparc64/kernel/setup.c
@@ -1,4 +1,4 @@
-/*  $Id: setup.c,v 1.5 1997/04/04 00:49:52 davem Exp $
+/*  $Id: setup.c,v 1.6 1997/05/04 07:21:04 davem Exp $
  *  linux/arch/sparc64/kernel/setup.c
  *
  *  Copyright (C) 1995,1996  David S. Miller (davem@caip.rutgers.edu)
@@ -261,6 +261,8 @@ __initfunc(void setup_arch(char **cmdline_p,
 	*cmdline_p = prom_getbootargs();
 	strcpy(saved_command_line, *cmdline_p);
 
+	prom_printf("BOOT: args[%s] saved[%s]\n", *cmdline_p, saved_command_line);
+
 	printk("ARCH: SUN4U\n");
 
 	boot_flags_init(*cmdline_p);
diff --git a/arch/sparc64/kernel/signal.c b/arch/sparc64/kernel/signal.c
new file mode 100644
index 000000000..1a595491a
--- /dev/null
+++ b/arch/sparc64/kernel/signal.c
@@ -0,0 +1,438 @@
+/*  $Id: signal.c,v 1.2 1997/05/18 08:42:15 davem Exp $
+ *  arch/sparc64/kernel/signal.c
+ *
+ *  Copyright (C) 1991, 1992  Linus Torvalds
+ *  Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
+ *  Copyright (C) 1996 Miguel de Icaza (miguel@nuclecu.unam.mx)
+ *  Copyright (C) 1997 Eddie C. Dost   (ecd@skynet.be)
+ *  Copyright (C) 1997 Jakub Jelinek   (jj@sunsite.mff.cuni.cz)
+ */
+
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/signal.h>
+#include <linux/errno.h>
+#include <linux/wait.h>
+#include <linux/ptrace.h>
+#include <linux/unistd.h>
+#include <linux/mm.h>
+
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <asm/ptrace.h>
+#include <asm/svr4.h>
+#include <asm/pgtable.h>
+#include <asm/fpumacro.h>
+#include <asm/smp_lock.h>
+
+#define _S(nr) (1<<((nr)-1))
+
+#define _BLOCKABLE (~(_S(SIGKILL) | _S(SIGSTOP)))
+
+asmlinkage int sys_wait4(pid_t pid, unsigned long *stat_addr,
+			 int options, unsigned long *ru);
+
+asmlinkage int do_signal(unsigned long oldmask, struct pt_regs * regs,
+			 unsigned long orig_o0, int ret_from_syscall);
+
+/* This turned off for production... */
+/* #define DEBUG_SIGNALS 1 */
+
+/* 
+ * The new signal frame, intended to be used for Linux applications only
+ * (we have enough in there to work with clone).
+ * All the interesting bits are in the info field.
+ */
+
+struct new_signal_frame {
+	struct sparc_stackf	ss;
+	__siginfo_t		info;
+	__siginfo_fpu_t *	fpu_save;
+	unsigned int		insns [2];
+	__siginfo_fpu_t		fpu_state;
+};
+
+/* Align macros */
+#define NF_ALIGNEDSZ  (((sizeof(struct new_signal_frame) + 7) & (~7)))
+
+/*
+ * atomically swap in the new signal mask, and wait for a signal.
+ * This is really tricky on the Sparc, watch out...
+ */
+asmlinkage void _sigpause_common(unsigned int set, struct pt_regs *regs)
+{
+	unsigned int mask;
+
+#ifdef CONFIG_SPARC32_COMPAT
+	if (current->tss.flags & SPARC_FLAG_32BIT) {
+		extern asmlinkage void _sigpause32_common(unsigned int, struct pt_regs *);
+		_sigpause32_common(set, regs);
+		return;
+	}
+#endif
+	spin_lock_irq(&current->sigmask_lock);
+	mask = current->blocked;
+	current->blocked = set & _BLOCKABLE;
+	spin_unlock_irq(&current->sigmask_lock);
+	
+	regs->tpc = regs->tnpc;
+	regs->tnpc += 4;
+
+	/* Condition codes and return value where set here for sigpause,
+	 * and so got used by setup_frame, which again causes sigreturn()
+	 * to return -EINTR.
+	 */
+	while (1) {
+		current->state = TASK_INTERRUPTIBLE;
+		schedule();
+		/*
+		 * Return -EINTR and set condition code here,
+		 * so the interrupted system call actually returns
+		 * these.
+		 */
+		regs->tstate |= (TSTATE_ICARRY|TSTATE_XCARRY);
+		regs->u_regs[UREG_I0] = EINTR;
+		if (do_signal(mask, regs, 0, 0))
+			return;
+	}
+}
+
+asmlinkage void do_sigpause(unsigned int set, struct pt_regs *regs)
+{
+	_sigpause_common(set, regs);
+}
+
+asmlinkage void do_sigsuspend(struct pt_regs *regs)
+{
+	_sigpause_common(regs->u_regs[UREG_I0], regs);
+}
+
+
+static inline void
+restore_fpu_state(struct pt_regs *regs, __siginfo_fpu_t *fpu)
+{
+#ifdef __SMP__
+	if (current->flags & PF_USEDFPU)
+		regs->tstate &= ~(TSTATE_PEF);
+#else
+	if (current == last_task_used_math) {
+		last_task_used_math = 0;
+		regs->tstate &= ~(TSTATE_PEF);
+	}
+#endif
+	current->used_math = 1;
+	current->flags &= ~PF_USEDFPU;
+
+	copy_from_user(&current->tss.float_regs[0],
+		       &fpu->si_float_regs[0],
+		       (sizeof(unsigned int) * 64));
+	__get_user(current->tss.fsr, &fpu->si_fsr);
+}
+
+void do_sigreturn(struct pt_regs *regs)
+{
+	struct new_signal_frame *sf;
+	unsigned long tpc, tnpc, tstate;
+	__siginfo_fpu_t *fpu_save;
+	int mask;
+
+#ifdef CONFIG_SPARC32_COMPAT
+	if (current->tss.flags & SPARC_FLAG_32BIT) {
+		extern asmlinkage void do_sigreturn32(struct pt_regs *);
+		do_sigreturn32(regs);
+		return;
+	}
+#endif	
+	synchronize_user_stack ();
+	sf = (struct new_signal_frame *) regs->u_regs [UREG_FP];
+	/* 1. Make sure we are not getting garbage from the user */
+	if (verify_area (VERIFY_READ, sf, sizeof (*sf))){
+		goto segv;
+	}
+	if (((unsigned long) sf) & 3){
+		goto segv;
+	}
+	get_user(tpc, &sf->info.si_regs.tpc);
+	__get_user(tnpc, &sf->info.si_regs.tnpc);
+	if ((tpc | tnpc) & 3){
+		goto segv;
+	}
+	regs->tpc = tpc;
+	regs->tnpc = tnpc;
+
+	/* 2. Restore the state */
+	__get_user(regs->y, &sf->info.si_regs.y);
+	__get_user(tstate, &sf->info.si_regs.tstate);
+	copy_from_user(regs->u_regs, sf->info.si_regs.u_regs, sizeof(regs->u_regs));
+
+	/* User can only change condition codes and FPU enabling in %tstate. */
+	regs->tstate &= ~(TSTATE_ICC | TSTATE_PEF);
+	regs->tstate |= (tstate & (TSTATE_ICC | TSTATE_PEF));
+
+	__get_user(fpu_save, &sf->fpu_save);
+	if (fpu_save)
+		restore_fpu_state(regs, &sf->fpu_state);
+	__get_user(mask, &sf->info.si_mask);
+	current->blocked = mask & _BLOCKABLE;
+	return;
+segv:
+	lock_kernel();
+	do_exit(SIGSEGV);
+}
+
+/* Checks if the fp is valid */
+static int invalid_frame_pointer(void *fp, int fplen)
+{
+	if ((((unsigned long) fp) & 7) || ((unsigned long)fp) > 0x80000000000ULL - fplen)
+		return 1;
+	return 0;
+}
+
+static inline void
+save_fpu_state(struct pt_regs *regs, __siginfo_fpu_t *fpu)
+{
+#ifdef __SMP__
+	if (current->flags & PF_USEDFPU) {
+		fpsave(&current->tss.float_regs[0], &current->tss.fsr);
+		regs->tstate &= ~(TSTATE_PEF);
+		current->flags &= ~(PF_USEDFPU);
+	}
+#else
+	if (current == last_task_used_math) {
+		fpsave((unsigned long *)&current->tss.float_regs[0], &current->tss.fsr);
+		last_task_used_math = 0;
+		regs->tstate &= ~(TSTATE_PEF);
+	}
+#endif
+	copy_to_user(&fpu->si_float_regs[0], &current->tss.float_regs[0],
+		     (sizeof(unsigned int) * 64));
+	__put_user(current->tss.fsr, &fpu->si_fsr);
+	current->used_math = 0;
+}
+
+static inline void
+new_setup_frame(struct sigaction *sa, struct pt_regs *regs,
+		  int signo, unsigned long oldmask)
+{
+	struct new_signal_frame *sf;
+	int sigframe_size;
+	unsigned long tmp;
+	int i;
+
+	/* 1. Make sure everything is clean */
+	synchronize_user_stack();
+	sigframe_size = NF_ALIGNEDSZ;
+	if (!current->used_math)
+		sigframe_size -= sizeof(__siginfo_fpu_t);
+
+	sf = (struct new_signal_frame *)(regs->u_regs[UREG_FP] - sigframe_size);
+	
+	if (invalid_frame_pointer (sf, sigframe_size)){
+		lock_kernel ();
+		do_exit(SIGILL);
+	}
+
+	if (current->tss.w_saved != 0){
+		printk ("%s[%d]: Invalid user stack frame for "
+			"signal delivery.\n", current->comm, current->pid);
+		lock_kernel ();
+		do_exit (SIGILL);
+	}
+
+	/* 2. Save the current process state */
+	copy_to_user(&sf->info.si_regs, regs, sizeof (*regs));
+
+	if (current->used_math) {
+		save_fpu_state(regs, &sf->fpu_state);
+		__put_user((u64)&sf->fpu_state, &sf->fpu_save);
+	} else {
+		__put_user(0, &sf->fpu_save);
+	}
+
+	__put_user(oldmask, &sf->info.si_mask);
+	for (i = 0; i < sizeof(struct reg_window)/8; i++) {
+		__get_user(tmp, (((u64 *)regs->u_regs[UREG_FP])+i));
+		__put_user(tmp, (((u64 *)sf)+i));
+	}
+	
+	/* 3. return to kernel instructions */
+	__put_user(0x821020d8, &sf->insns[0]); /* mov __NR_sigreturn, %g1 */
+	__put_user(0x91d02011, &sf->insns[1]); /* t 0x11 */
+
+	/* 4. signal handler back-trampoline and parameters */
+	regs->u_regs[UREG_FP] = (unsigned long) sf;
+	regs->u_regs[UREG_I0] = signo;
+	regs->u_regs[UREG_I1] = (unsigned long) &sf->info;
+	regs->u_regs[UREG_I7] = (unsigned long) (&(sf->insns[0]) - 2);
+
+	/* 5. signal handler */
+	regs->tpc = (unsigned long) sa->sa_handler;
+	regs->tnpc = (regs->tpc + 4);
+
+	/* Flush instruction space. */
+	__asm__ __volatile__ ("flush %0; flush %0 + 4" : : "r" (&(sf->insns[0])));
+		
+}
+
+static inline void handle_signal(unsigned long signr, struct sigaction *sa,
+				 unsigned long oldmask, struct pt_regs *regs)
+{
+	new_setup_frame(sa, regs, signr, oldmask);
+	if(sa->sa_flags & SA_ONESHOT)
+		sa->sa_handler = NULL;
+	if(!(sa->sa_flags & SA_NOMASK))
+		current->blocked |= (sa->sa_mask | _S(signr)) & _BLOCKABLE;
+}
+
+static inline void syscall_restart(unsigned long orig_i0, struct pt_regs *regs,
+				     struct sigaction *sa)
+{
+	switch(regs->u_regs[UREG_I0]) {
+		case ERESTARTNOHAND:
+		no_system_call_restart:
+			regs->u_regs[UREG_I0] = EINTR;
+			regs->tstate |= (TSTATE_ICARRY|TSTATE_XCARRY);
+			break;
+		case ERESTARTSYS:
+			if(!(sa->sa_flags & SA_RESTART))
+				goto no_system_call_restart;
+		/* fallthrough */
+		case ERESTARTNOINTR:
+			regs->u_regs[UREG_I0] = orig_i0;
+			regs->tpc -= 4;
+			regs->tnpc -= 4;
+	}
+}
+
+/* Note that 'init' is a special process: it doesn't get signals it doesn't
+ * want to handle. Thus you cannot kill init even with a SIGKILL even by
+ * mistake.
+ */
+asmlinkage int do_signal(unsigned long oldmask, struct pt_regs * regs,
+			   unsigned long orig_i0, int restart_syscall)
+{
+	unsigned long signr, mask = ~current->blocked;
+	struct sigaction *sa;
+
+#ifdef CONFIG_SPARC32_COMPAT
+	if (current->tss.flags & SPARC_FLAG_32BIT) {
+		extern asmlinkage int do_signal32(unsigned long, struct pt_regs *,
+						  unsigned long, int);
+		return do_signal32(oldmask, regs, orig_i0, restart_syscall);
+	}
+#endif	
+	while ((signr = current->signal & mask) != 0) {
+		signr = ffz(~signr);
+		clear_bit(signr, &current->signal);
+		sa = current->sig->action + signr;
+		signr++;
+		if ((current->flags & PF_PTRACED) && signr != SIGKILL) {
+			current->exit_code = signr;
+			current->state = TASK_STOPPED;
+			notify_parent(current);
+			schedule();
+			if (!(signr = current->exit_code))
+				continue;
+			current->exit_code = 0;
+			if (signr == SIGSTOP)
+				continue;
+			if (_S(signr) & current->blocked) {
+				current->signal |= _S(signr);
+				continue;
+			}
+			sa = current->sig->action + signr - 1;
+		}
+		if(sa->sa_handler == SIG_IGN) {
+			if(signr != SIGCHLD)
+				continue;
+
+                        /* sys_wait4() grabs the master kernel lock, so
+                         * we need not do so, that sucker should be
+                         * threaded and would not be that difficult to
+                         * do anyways.
+                         */
+                        while(sys_wait4(-1, NULL, WNOHANG, NULL) > 0)
+                                ;
+			continue;
+		}
+		if(sa->sa_handler == SIG_DFL) {
+			if(current->pid == 1)
+				continue;
+			switch(signr) {
+			case SIGCONT: case SIGCHLD: case SIGWINCH:
+				continue;
+
+			case SIGTSTP: case SIGTTIN: case SIGTTOU:
+				if (is_orphaned_pgrp(current->pgrp))
+					continue;
+
+			case SIGSTOP:
+				if (current->flags & PF_PTRACED)
+					continue;
+				current->state = TASK_STOPPED;
+				current->exit_code = signr;
+				if(!(current->p_pptr->sig->action[SIGCHLD-1].sa_flags &
+				     SA_NOCLDSTOP))
+					notify_parent(current);
+				schedule();
+				continue;
+
+			case SIGQUIT: case SIGILL: case SIGTRAP:
+			case SIGABRT: case SIGFPE: case SIGSEGV: case SIGBUS:
+				if(current->binfmt && current->binfmt->core_dump) {
+					if(current->binfmt->core_dump(signr, regs))
+						signr |= 0x80;
+				}
+#ifdef DEBUG_SIGNALS
+				/* Very useful to debug dynamic linker problems */
+				printk ("Sig ILL going...\n");
+				show_regs (regs);
+#endif
+				/* fall through */
+			default:
+				current->signal |= _S(signr & 0x7f);
+				current->flags |= PF_SIGNALED;
+				do_exit(signr);
+			}
+		}
+		if(restart_syscall)
+			syscall_restart(orig_i0, regs, sa);
+		handle_signal(signr, sa, oldmask, regs);
+		return 1;
+	}
+	if(restart_syscall &&
+	   (regs->u_regs[UREG_I0] == ERESTARTNOHAND ||
+	    regs->u_regs[UREG_I0] == ERESTARTSYS ||
+	    regs->u_regs[UREG_I0] == ERESTARTNOINTR)) {
+		/* replay the system call when we are done */
+		regs->u_regs[UREG_I0] = orig_i0;
+		regs->tpc -= 4;
+		regs->tnpc -= 4;
+	}
+	return 0;
+}
+
+asmlinkage int
+sys_sigstack(struct sigstack *ssptr, struct sigstack *ossptr)
+{
+	int ret = -EFAULT;
+
+	lock_kernel();
+	/* First see if old state is wanted. */
+	if(ossptr) {
+		if (put_user ((u64)current->tss.sstk_info.the_stack, &ossptr->the_stack) ||
+		    __put_user (current->tss.sstk_info.cur_status, &ossptr->cur_status))
+			goto out;
+	}
+
+	/* Now see if we want to update the new state. */
+	if(ssptr) {
+		if (get_user ((u64)current->tss.sstk_info.the_stack, &ssptr->the_stack) ||
+		    __put_user (current->tss.sstk_info.cur_status, &ssptr->cur_status))
+			goto out;
+	}
+	ret = 0;
+out:
+	unlock_kernel();
+	return ret;
+}
diff --git a/arch/sparc64/kernel/signal32.c b/arch/sparc64/kernel/signal32.c
index e1129dfd6..d8e61511d 100644
--- a/arch/sparc64/kernel/signal32.c
+++ b/arch/sparc64/kernel/signal32.c
@@ -1,4 +1,4 @@
-/*  $Id: signal32.c,v 1.6 1997/04/16 10:27:17 jj Exp $
+/*  $Id: signal32.c,v 1.8 1997/05/18 08:42:15 davem Exp $
  *  arch/sparc64/kernel/signal32.c
  *
  *  Copyright (C) 1991, 1992  Linus Torvalds
@@ -30,8 +30,6 @@
 
 #define _BLOCKABLE (~(_S(SIGKILL) | _S(SIGSTOP)))
 
-#define synchronize_user_stack() do { } while (0)
-
 asmlinkage int sys_wait4(pid_t pid, unsigned long *stat_addr,
 			 int options, unsigned long *ru);
 
@@ -117,17 +115,6 @@ asmlinkage void _sigpause32_common(unsigned int set, struct pt_regs *regs)
 	}
 }
 
-asmlinkage void do_sigpause32(unsigned int set, struct pt_regs *regs)
-{
-	_sigpause32_common(set, regs);
-}
-
-asmlinkage void do_sigsuspend32(struct pt_regs *regs)
-{
-	_sigpause32_common(regs->u_regs[UREG_I0], regs);
-}
-
-
 static inline void
 restore_fpu_state32(struct pt_regs *regs, __siginfo_fpu32_t *fpu)
 {
@@ -248,7 +235,9 @@ setup_frame32(struct sigaction *sa, unsigned long pc, unsigned long npc,
 {
 	struct signal_sframe32 *sframep;
 	struct sigcontext32 *sc;
+#if 0	
 	int window = 0;
+#endif	
 	int old_status = current->tss.sstk_info.cur_status;
 	unsigned psr;
 	int i;
@@ -424,7 +413,9 @@ setup_svr4_frame32(struct sigaction *sa, unsigned long pc, unsigned long npc,
 	svr4_mcontext_t *mc;
 	svr4_gwindows_t *gw;
 	svr4_ucontext_t *uc;
+#if 0	
 	int window = 0;
+#endif	
 	unsigned psr;
 	int i;
 
diff --git a/arch/sparc64/kernel/sparcelf32.c b/arch/sparc64/kernel/sparcelf32.c
deleted file mode 100644
index 855c636e4..000000000
--- a/arch/sparc64/kernel/sparcelf32.c
+++ /dev/null
@@ -1,1281 +0,0 @@
-/* sparcelf32.c: Support 32-bit Sparc ELF binaries on Ultra.
- *
- * This is just binfmt_elf.c with hooks so that the types are those
- * for a 32-bit ELF binaries.
- */
-
-/* This makes it work. */
-#define ELF_ARCH	EM_SPARC
-#define ELF_CLASS	ELFCLASS32
-#define ELF_DATA	ELFDATA2MSB;
-
-/*
- * linux/fs/binfmt_elf.c
- *
- * These are the functions used to load ELF format executables as used
- * on SVr4 machines.  Information on the format may be found in the book
- * "UNIX SYSTEM V RELEASE 4 Programmers Guide: Ansi C and Programming Support
- * Tools".
- *
- * Copyright 1993, 1994: Eric Youngdale (ericy@cais.com).
- */
-
-#include <linux/module.h>
-
-#include <linux/fs.h>
-#include <linux/stat.h>
-#include <linux/sched.h>
-#include <linux/mm.h>
-#include <linux/mman.h>
-#include <linux/a.out.h>
-#include <linux/errno.h>
-#include <linux/signal.h>
-#include <linux/binfmts.h>
-#include <linux/string.h>
-#include <linux/fcntl.h>
-#include <linux/ptrace.h>
-#include <linux/malloc.h>
-#include <linux/shm.h>
-#include <linux/personality.h>
-#include <linux/elfcore.h>
-#include <linux/init.h>
-
-#include <asm/uaccess.h>
-#include <asm/pgtable.h>
-
-#include <linux/config.h>
-
-#define DLINFO_ITEMS 12
-
-#include <linux/elf.h>
-
-static int load_elf32_binary(struct linux_binprm * bprm, struct pt_regs * regs);
-static int load_elf32_library(int fd);
-extern int dump_fpu (struct pt_regs *, elf_fpregset_t *);
-extern void dump_thread(struct pt_regs *, struct user *);
-
-extern unsigned long get_unmapped_area(unsigned long addr, unsigned long len);
-
-/*
- * If we don't support core dumping, then supply a NULL so we
- * don't even try.
- */
-#ifdef USE_ELF_CORE_DUMP
-static int elf32_core_dump(long signr, struct pt_regs * regs);
-#else
-#define elf32_core_dump	NULL
-#endif
-
-#define ELF_PAGESTART(_v) ((_v) & ~(unsigned long)(ELF_EXEC_PAGESIZE-1))
-#define ELF_PAGEOFFSET(_v) ((_v) & (ELF_EXEC_PAGESIZE-1))
-
-static struct linux_binfmt elf32_format = {
-#ifndef MODULE
-	NULL, NULL, load_elf32_binary, load_elf32_library, elf32_core_dump
-#else
-	NULL, &__this_module, load_elf32_binary, load_elf32_library, elf32_core_dump
-#endif
-};
-
-static void set_brk(unsigned long start, unsigned long end)
-{
-	start = PAGE_ALIGN(start);
-	end = PAGE_ALIGN(end);
-	if (end <= start)
-		return;
-	do_mmap(NULL, start, end - start,
-		PROT_READ | PROT_WRITE | PROT_EXEC,
-		MAP_FIXED | MAP_PRIVATE, 0);
-}
-
-
-/* We need to explicitly zero any fractional pages
-   after the data section (i.e. bss).  This would
-   contain the junk from the file that should not
-   be in memory */
-
-
-static void padzero(unsigned long elf_bss)
-{
-	unsigned long nbyte;
-
-	nbyte = elf_bss & (PAGE_SIZE-1);
-	if (nbyte) {
-		nbyte = PAGE_SIZE - nbyte;
-		clear_user((void *) elf_bss, nbyte);
-	}
-}
-
-__u32 *create_elf32_tables(char *p, int argc, int envc,
-			   struct elfhdr * exec,
-			   unsigned long load_addr,
-			   unsigned long interp_load_addr, int ibcs)
-{
-	__u32 *argv, *envp;
-	__u32 *sp, *csp;
-
-	/*
-	 * Force 16 byte alignment here for generality.
-	 */
-	sp = (__u32 *) (~15UL & (unsigned long) p);
-	csp = sp;
-	csp -= exec ? DLINFO_ITEMS*2 : 2;
-	csp -= envc+1;
-	csp -= argc+1;
-	if (!(((unsigned long) csp) & 4))
-		sp--;
-
-	/*
-	 * Put the ELF interpreter info on the stack
-	 */
-#define NEW_AUX_ENT(nr, id, val) \
-	  __put_user ((id), sp+(nr*2)); \
-	  __put_user ((val), sp+(nr*2+1)); \
-
-	sp -= 2;
-	NEW_AUX_ENT(0, AT_NULL, 0);
-
-	if (exec) {
-		sp -= 11*2;
-
-	  NEW_AUX_ENT (0, AT_PHDR, load_addr + exec->e_phoff);
-	  NEW_AUX_ENT (1, AT_PHENT, sizeof (struct elf_phdr));
-	  NEW_AUX_ENT (2, AT_PHNUM, exec->e_phnum);
-	  NEW_AUX_ENT (3, AT_PAGESZ, PAGE_SIZE);
-	  NEW_AUX_ENT (4, AT_BASE, interp_load_addr);
-	  NEW_AUX_ENT (5, AT_FLAGS, 0);
-	  NEW_AUX_ENT (6, AT_ENTRY, (__u32) exec->e_entry);
-	  NEW_AUX_ENT (7, AT_UID, (__u32) current->uid);
-	  NEW_AUX_ENT (8, AT_EUID, (__u32) current->euid);
-	  NEW_AUX_ENT (9, AT_GID, (__u32) current->gid);
-	  NEW_AUX_ENT (10, AT_EGID, (__u32) current->egid);
-	}
-#undef NEW_AUX_ENT
-
-	sp -= envc+1;
-	envp = (__u32 *) sp;
-	sp -= argc+1;
-	argv = (__u32 *) sp;
-	if (!ibcs) {
-		__put_user(((__u32)(long) envp),--sp);
-		__put_user(((__u32)(long) argv),--sp);
-	}
-
-	__put_user((__u32)argc,--sp);
-	current->mm->arg_start = (unsigned long) p;
-	while (argc-->0) {
-		__put_user(((__u32)(long)p),argv++);
-		p += strlen_user(p);
-	}
-	__put_user(NULL, argv);
-	current->mm->arg_end = current->mm->env_start = (unsigned long) p;
-	while (envc-->0) {
-		__put_user(((__u32)(long)p),envp++);
-		p += strlen_user(p);
-	}
-	__put_user(NULL, envp);
-	current->mm->env_end = (unsigned long) p;
-	return sp;
-}
-
-
-/* This is much more generalized than the library routine read function,
-   so we keep this separate.  Technically the library read function
-   is only provided so that we can read a.out libraries that have
-   an ELF header */
-
-static unsigned long load_elf32_interp(struct elfhdr * interp_elf_ex,
-				       struct inode * interpreter_inode,
-				       unsigned long *interp_load_addr)
-{
-	struct file * file;
-	struct elf_phdr *elf_phdata  =  NULL;
-	struct elf_phdr *eppnt;
-	unsigned long load_addr;
-	int load_addr_set = 0;
-	int elf_exec_fileno;
-	int retval;
-	unsigned long last_bss, elf_bss;
-	unsigned long error;
-	int i;
-
-	elf_bss = 0;
-	last_bss = 0;
-	error = load_addr = 0;
-
-	/* First of all, some simple consistency checks */
-	if ((interp_elf_ex->e_type != ET_EXEC &&
-	    interp_elf_ex->e_type != ET_DYN) ||
-	   !elf_check_arch(interp_elf_ex->e_machine) ||
-	   (!interpreter_inode->i_op ||
-	    !interpreter_inode->i_op->default_file_ops->mmap)){
-		return ~0UL;
-	}
-
-	/* Now read in all of the header information */
-
-	if (sizeof(struct elf_phdr) * interp_elf_ex->e_phnum > PAGE_SIZE)
-	    return ~0UL;
-
-	elf_phdata =  (struct elf_phdr *)
-		kmalloc(sizeof(struct elf_phdr) * interp_elf_ex->e_phnum,
-			GFP_KERNEL);
-	if (!elf_phdata)
-	  return ~0UL;
-
-	/*
-	 * If the size of this structure has changed, then punt, since
-	 * we will be doing the wrong thing.
-	 */
-	if (interp_elf_ex->e_phentsize != sizeof(struct elf_phdr))
-	  {
-	    kfree(elf_phdata);
-	    return ~0UL;
-	  }
-
-	retval = read_exec(interpreter_inode, interp_elf_ex->e_phoff,
-			   (char *) elf_phdata,
-			   sizeof(struct elf_phdr) * interp_elf_ex->e_phnum, 1);
-
-	if (retval < 0) {
-		kfree (elf_phdata);
-		return retval;
- 	}
-
-	elf_exec_fileno = open_inode(interpreter_inode, O_RDONLY);
-	if (elf_exec_fileno < 0) {
-	  kfree(elf_phdata);
-	  return ~0UL;
-	}
-
-	file = current->files->fd[elf_exec_fileno];
-
-	eppnt = elf_phdata;
-	for(i=0; i<interp_elf_ex->e_phnum; i++, eppnt++)
-	  if (eppnt->p_type == PT_LOAD) {
-	    int elf_type = MAP_PRIVATE | MAP_DENYWRITE;
-	    int elf_prot = 0;
-	    unsigned long vaddr = 0;
-	    unsigned long k;
-
-	    if (eppnt->p_flags & PF_R) elf_prot =  PROT_READ;
-	    if (eppnt->p_flags & PF_W) elf_prot |= PROT_WRITE;
-	    if (eppnt->p_flags & PF_X) elf_prot |= PROT_EXEC;
-	    if (interp_elf_ex->e_type == ET_EXEC || load_addr_set) {
-	    	elf_type |= MAP_FIXED;
-	    	vaddr = eppnt->p_vaddr;
-	    } else {
-		load_addr = get_unmapped_area(0, eppnt->p_filesz +
-					ELF_PAGEOFFSET(eppnt->p_vaddr));
-	    }
-
-	    error = do_mmap(file,
-			    load_addr + ELF_PAGESTART(vaddr),
-			    eppnt->p_filesz + ELF_PAGEOFFSET(eppnt->p_vaddr),
-			    elf_prot,
-			    elf_type,
-			    eppnt->p_offset - ELF_PAGEOFFSET(eppnt->p_vaddr));
-
-	    if (error > -1024UL) {
-	      /* Real error */
-	      sys_close(elf_exec_fileno);
-	      kfree(elf_phdata);
-	      return ~0UL;
-	    }
-
-	    if (!load_addr_set && interp_elf_ex->e_type == ET_DYN) {
-	      load_addr = error;
-	      load_addr_set = 1;
-	    }
-
-	    /*
-	     * Find the end of the file  mapping for this phdr, and keep
-	     * track of the largest address we see for this.
-	     */
-	    k = load_addr + eppnt->p_vaddr + eppnt->p_filesz;
-	    if (k > elf_bss) elf_bss = k;
-
-	    /*
-	     * Do the same thing for the memory mapping - between
-	     * elf_bss and last_bss is the bss section.
-	     */
-	    k = load_addr + eppnt->p_memsz + eppnt->p_vaddr;
-	    if (k > last_bss) last_bss = k;
-	  }
-
-	/* Now use mmap to map the library into memory. */
-
-	sys_close(elf_exec_fileno);
-
-	/*
-	 * Now fill out the bss section.  First pad the last page up
-	 * to the page boundary, and then perform a mmap to make sure
-	 * that there are zeromapped pages up to and including the last
-	 * bss page.
-	 */
-	padzero(elf_bss);
-	elf_bss = ELF_PAGESTART(elf_bss + ELF_EXEC_PAGESIZE - 1); /* What we have mapped so far */
-
-	/* Map the last of the bss segment */
-	if (last_bss > elf_bss)
-	  do_mmap(NULL, elf_bss, last_bss-elf_bss,
-		  PROT_READ|PROT_WRITE|PROT_EXEC,
-		  MAP_FIXED|MAP_PRIVATE, 0);
-	kfree(elf_phdata);
-
-	*interp_load_addr = load_addr;
-	return ((unsigned long) interp_elf_ex->e_entry) + load_addr;
-}
-
-static unsigned long load_aout32_interp(struct exec * interp_ex,
-			     struct inode * interpreter_inode)
-{
-  int retval;
-  unsigned long elf_entry;
-
-  current->mm->brk = interp_ex->a_bss +
-    (current->mm->end_data = interp_ex->a_data +
-     (current->mm->end_code = interp_ex->a_text));
-  elf_entry = interp_ex->a_entry;
-
-
-  if (N_MAGIC(*interp_ex) == OMAGIC) {
-    do_mmap(NULL, 0, interp_ex->a_text+interp_ex->a_data,
-	    PROT_READ|PROT_WRITE|PROT_EXEC,
-	    MAP_FIXED|MAP_PRIVATE, 0);
-    retval = read_exec(interpreter_inode, 32, (char *) 0,
-		       interp_ex->a_text+interp_ex->a_data, 0);
-  } else if (N_MAGIC(*interp_ex) == ZMAGIC || N_MAGIC(*interp_ex) == QMAGIC) {
-    do_mmap(NULL, 0, interp_ex->a_text+interp_ex->a_data,
-	    PROT_READ|PROT_WRITE|PROT_EXEC,
-	    MAP_FIXED|MAP_PRIVATE, 0);
-    retval = read_exec(interpreter_inode,
-		       N_TXTOFF(*interp_ex) ,
-		       (char *) N_TXTADDR(*interp_ex),
-		       interp_ex->a_text+interp_ex->a_data, 0);
-  } else
-    retval = -1;
-
-  if (retval >= 0)
-    do_mmap(NULL, ELF_PAGESTART(interp_ex->a_text + interp_ex->a_data + ELF_EXEC_PAGESIZE - 1),
-	    interp_ex->a_bss,
-	    PROT_READ|PROT_WRITE|PROT_EXEC,
-	    MAP_FIXED|MAP_PRIVATE, 0);
-  if (retval < 0) return ~0UL;
-  return elf_entry;
-}
-
-/*
- * These are the functions used to load ELF style executables and shared
- * libraries.  There is no binary dependent code anywhere else.
- */
-
-#define INTERPRETER_NONE 0
-#define INTERPRETER_AOUT 1
-#define INTERPRETER_ELF 2
-
-
-static inline int
-do_load_elf32_binary(struct linux_binprm * bprm, struct pt_regs * regs)
-{
-	struct elfhdr elf_ex;
-	struct elfhdr interp_elf_ex;
-	struct file * file;
-  	struct exec interp_ex;
-	struct inode *interpreter_inode;
-	unsigned long load_addr;
-	int load_addr_set = 0;
-	unsigned int interpreter_type = INTERPRETER_NONE;
-	unsigned char ibcs2_interpreter;
-	int i;
-	int old_fs;
-	int error;
-	struct elf_phdr * elf_ppnt, *elf_phdata;
-	int elf_exec_fileno;
-	unsigned long elf_bss, k, elf_brk;
-	int retval;
-	char * elf_interpreter;
-	unsigned long elf_entry, interp_load_addr = 0;
-	int status;
-	unsigned long start_code, end_code, end_data;
-	unsigned long elf_stack;
-	char passed_fileno[6];
-
-	ibcs2_interpreter = 0;
-	status = 0;
-	load_addr = 0;
-	elf_ex = *((struct elfhdr *) bprm->buf);	  /* exec-header */
-
-	if (elf_ex.e_ident[0] != 0x7f ||
-	    strncmp(&elf_ex.e_ident[1], "ELF",3) != 0) {
-		return  -ENOEXEC;
-	}
-
-
-	/* First of all, some simple consistency checks */
-	if ((elf_ex.e_type != ET_EXEC &&
-	    elf_ex.e_type != ET_DYN) ||
-	   (! elf_check_arch(elf_ex.e_machine)) ||
-	   (!bprm->inode->i_op || !bprm->inode->i_op->default_file_ops ||
-	    !bprm->inode->i_op->default_file_ops->mmap)){
-		return -ENOEXEC;
-	}
-
-	/* Now read in all of the header information */
-
-	elf_phdata = (struct elf_phdr *) kmalloc(elf_ex.e_phentsize *
-						 elf_ex.e_phnum, GFP_KERNEL);
-	if (elf_phdata == NULL) {
-		return -ENOMEM;
-	}
-
-	retval = read_exec(bprm->inode, elf_ex.e_phoff, (char *) elf_phdata,
-			   elf_ex.e_phentsize * elf_ex.e_phnum, 1);
-	if (retval < 0) {
-		kfree (elf_phdata);
-		return retval;
-	}
-
-	elf_ppnt = elf_phdata;
-
-	elf_bss = 0;
-	elf_brk = 0;
-
-	elf_exec_fileno = open_inode(bprm->inode, O_RDONLY);
-
-	if (elf_exec_fileno < 0) {
-		kfree (elf_phdata);
-		return elf_exec_fileno;
-	}
-
-	file = current->files->fd[elf_exec_fileno];
-
-	elf_stack = ~0UL;
-	elf_interpreter = NULL;
-	start_code = ~0UL;
-	end_code = 0;
-	end_data = 0;
-
-	for(i=0;i < elf_ex.e_phnum; i++){
-		if (elf_ppnt->p_type == PT_INTERP) {
-		  	if ( elf_interpreter != NULL )
-			{
-				kfree (elf_phdata);
-				kfree(elf_interpreter);
-				sys_close(elf_exec_fileno);
-				return -EINVAL;
-			}
-
-			/* This is the program interpreter used for
-			 * shared libraries - for now assume that this
-			 * is an a.out format binary
-			 */
-
-			elf_interpreter = (char *) kmalloc(elf_ppnt->p_filesz,
-							   GFP_KERNEL);
-			if (elf_interpreter == NULL) {
-				kfree (elf_phdata);
-				sys_close(elf_exec_fileno);
-				return -ENOMEM;
-			}
-
-			retval = read_exec(bprm->inode,elf_ppnt->p_offset,
-					   elf_interpreter,
-					   elf_ppnt->p_filesz, 1);
-			/* If the program interpreter is one of these two,
-			   then assume an iBCS2 image. Otherwise assume
-			   a native linux image. */
-			if (strcmp(elf_interpreter,"/usr/lib/libc.so.1") == 0 ||
-			    strcmp(elf_interpreter,"/usr/lib/ld.so.1") == 0)
-			  ibcs2_interpreter = 1;
-#if 0
-			printk("Using ELF interpreter %s\n", elf_interpreter);
-#endif
-			if (retval >= 0) {
-				old_fs = get_fs(); /* This could probably be optimized */
-				set_fs(get_ds());
-				retval = open_namei(elf_interpreter, 0, 0,
-						    &interpreter_inode, NULL);
-				set_fs(old_fs);
-			}
-
-			if (retval >= 0)
-				retval = read_exec(interpreter_inode,0,bprm->buf,128, 1);
-
-			if (retval >= 0) {
-				interp_ex = *((struct exec *) bprm->buf);		/* exec-header */
-				interp_elf_ex = *((struct elfhdr *) bprm->buf);	  /* exec-header */
-
-			}
-			if (retval < 0) {
-				kfree (elf_phdata);
-				kfree(elf_interpreter);
-				sys_close(elf_exec_fileno);
-				return retval;
-			}
-		}
-		elf_ppnt++;
-	}
-
-	/* Some simple consistency checks for the interpreter */
-	if (elf_interpreter){
-		interpreter_type = INTERPRETER_ELF | INTERPRETER_AOUT;
-
-		/* Now figure out which format our binary is */
-		if ((N_MAGIC(interp_ex) != OMAGIC) &&
-		    (N_MAGIC(interp_ex) != ZMAGIC) &&
-		    (N_MAGIC(interp_ex) != QMAGIC))
-		  interpreter_type = INTERPRETER_ELF;
-
-		if (interp_elf_ex.e_ident[0] != 0x7f ||
-		    strncmp(&interp_elf_ex.e_ident[1], "ELF",3) != 0)
-		  interpreter_type &= ~INTERPRETER_ELF;
-
-		if (!interpreter_type)
-		  {
-		    kfree(elf_interpreter);
-		    kfree(elf_phdata);
-		    sys_close(elf_exec_fileno);
-		    return -ELIBBAD;
-		  }
-	}
-
-	/* OK, we are done with that, now set up the arg stuff,
-	   and then start this sucker up */
-
-	if (!bprm->sh_bang) {
-		char * passed_p;
-
-		if (interpreter_type == INTERPRETER_AOUT) {
-		  sprintf(passed_fileno, "%d", elf_exec_fileno);
-		  passed_p = passed_fileno;
-
-		  if (elf_interpreter) {
-		    bprm->p = copy_strings(1,&passed_p,bprm->page,bprm->p,2);
-		    bprm->argc++;
-		  }
-		}
-		if (!bprm->p) {
-			if (elf_interpreter) {
-			      kfree(elf_interpreter);
-			}
-			kfree (elf_phdata);
-			sys_close(elf_exec_fileno);
-			return -E2BIG;
-		}
-	}
-
-	/* OK, This is the point of no return */
-	flush_old_exec(bprm);
-
-	current->mm->end_data = 0;
-	current->mm->end_code = 0;
-	current->mm->start_mmap = ELF_START_MMAP;
-	current->mm->mmap = NULL;
-	elf_entry = (unsigned long) elf_ex.e_entry;
-
-	/* Do this so that we can load the interpreter, if need be.  We will
-	   change some of these later */
-	current->mm->rss = 0;
-	bprm->p = setup_arg_pages(bprm->p, bprm);
-	current->mm->start_stack = bprm->p;
-
-	/* Now we do a little grungy work by mmaping the ELF image into
-	   the correct location in memory.  At this point, we assume that
-	   the image should be loaded at fixed address, not at a variable
-	   address. */
-
-	old_fs = get_fs();
-	set_fs(get_ds());
-	for(i = 0, elf_ppnt = elf_phdata; i < elf_ex.e_phnum; i++, elf_ppnt++) {
-		if (elf_ppnt->p_type == PT_LOAD) {
-			int elf_prot = 0;
-			if (elf_ppnt->p_flags & PF_R) elf_prot |= PROT_READ;
-			if (elf_ppnt->p_flags & PF_W) elf_prot |= PROT_WRITE;
-			if (elf_ppnt->p_flags & PF_X) elf_prot |= PROT_EXEC;
-
-			error = do_mmap(file,
-					ELF_PAGESTART(elf_ppnt->p_vaddr),
-					(elf_ppnt->p_filesz +
-					 ELF_PAGEOFFSET(elf_ppnt->p_vaddr)),
-					elf_prot,
-					(MAP_FIXED | MAP_PRIVATE |
-					 MAP_DENYWRITE | MAP_EXECUTABLE),
-					(elf_ppnt->p_offset -
-					 ELF_PAGEOFFSET(elf_ppnt->p_vaddr)));
-
-#ifdef LOW_ELF_STACK
-			if (ELF_PAGESTART(elf_ppnt->p_vaddr) < elf_stack)
-				elf_stack = ELF_PAGESTART(elf_ppnt->p_vaddr);
-#endif
-
-			if (!load_addr_set) {
-			  load_addr = elf_ppnt->p_vaddr - elf_ppnt->p_offset;
-			  load_addr_set = 1;
-			}
-			k = elf_ppnt->p_vaddr;
-			if (k < start_code) start_code = k;
-			k = elf_ppnt->p_vaddr + elf_ppnt->p_filesz;
-			if (k > elf_bss) elf_bss = k;
-#if 1
-			if ((elf_ppnt->p_flags & PF_X) && end_code <  k)
-#else
-			if ( !(elf_ppnt->p_flags & PF_W) && end_code <  k)
-#endif
-				end_code = k;
-			if (end_data < k) end_data = k;
-			k = elf_ppnt->p_vaddr + elf_ppnt->p_memsz;
-			if (k > elf_brk) elf_brk = k;
-		}
-	}
-	set_fs(old_fs);
-
-	if (elf_interpreter) {
-		if (interpreter_type & 1)
-			elf_entry = load_aout32_interp(&interp_ex,
-						       interpreter_inode);
-		else if (interpreter_type & 2)
-			elf_entry = load_elf32_interp(&interp_elf_ex,
-						      interpreter_inode,
-						      &interp_load_addr);
-
-		iput(interpreter_inode);
-		kfree(elf_interpreter);
-
-		if (elf_entry == ~0UL) {
-			printk("Unable to load interpreter\n");
-			kfree(elf_phdata);
-			send_sig(SIGSEGV, current, 0);
-			return 0;
-		}
-	}
-
-	kfree(elf_phdata);
-
-	if (interpreter_type != INTERPRETER_AOUT) sys_close(elf_exec_fileno);
-	current->personality = (ibcs2_interpreter ? PER_SVR4 : PER_LINUX);
-
-	if (current->exec_domain && current->exec_domain->module)
-		__MOD_DEC_USE_COUNT(current->exec_domain->module);
-	if (current->binfmt && current->binfmt->module)
-		__MOD_DEC_USE_COUNT(current->binfmt->module);
-	current->exec_domain = lookup_exec_domain(current->personality);
-	current->binfmt = &elf32_format;
-	if (current->exec_domain && current->exec_domain->module)
-		__MOD_INC_USE_COUNT(current->exec_domain->module);
-	if (current->binfmt && current->binfmt->module)
-		__MOD_INC_USE_COUNT(current->binfmt->module);
-
-#ifndef VM_STACK_FLAGS
-	current->executable = bprm->inode;
-	bprm->inode->i_count++;
-#endif
-#ifdef LOW_ELF_STACK
-	current->start_stack = bprm->p = elf_stack - 4;
-#endif
-	current->suid = current->euid = current->fsuid = bprm->e_uid;
-	current->sgid = current->egid = current->fsgid = bprm->e_gid;
-	current->flags &= ~PF_FORKNOEXEC;
-	bprm->p = (unsigned long)
-	  create_elf32_tables((char *)bprm->p,
-			      bprm->argc,
-			      bprm->envc,
-			      (interpreter_type == INTERPRETER_ELF ? &elf_ex : NULL),
-			      load_addr,
-			      interp_load_addr,
-			      (interpreter_type == INTERPRETER_AOUT ? 0 : 1));
-	if (interpreter_type == INTERPRETER_AOUT)
-	  current->mm->arg_start += strlen(passed_fileno) + 1;
-	current->mm->start_brk = current->mm->brk = elf_brk;
-	current->mm->end_code = end_code;
-	current->mm->start_code = start_code;
-	current->mm->end_data = end_data;
-	current->mm->start_stack = bprm->p;
-
-	/* Calling set_brk effectively mmaps the pages that we need for the bss and break
-	   sections */
-	set_brk(elf_bss, elf_brk);
-
-	padzero(elf_bss);
-
-#if 0
-	printk("(start_brk) %x\n" , current->mm->start_brk);
-	printk("(end_code) %x\n" , current->mm->end_code);
-	printk("(start_code) %x\n" , current->mm->start_code);
-	printk("(end_data) %x\n" , current->mm->end_data);
-	printk("(start_stack) %x\n" , current->mm->start_stack);
-	printk("(brk) %x\n" , current->mm->brk);
-#endif
-
-	if ( current->personality == PER_SVR4 )
-	{
-		/* Why this, you ask???  Well SVr4 maps page 0 as read-only,
-		   and some applications "depend" upon this behavior.
-		   Since we do not have the power to recompile these, we
-		   emulate the SVr4 behavior.  Sigh.  */
-		error = do_mmap(NULL, 0, 4096, PROT_READ | PROT_EXEC,
-				MAP_FIXED | MAP_PRIVATE, 0);
-	}
-
-#ifdef ELF_PLAT_INIT
-	/*
-	 * The ABI may specify that certain registers be set up in special
-	 * ways (on i386 %edx is the address of a DT_FINI function, for
-	 * example.  This macro performs whatever initialization to
-	 * the regs structure is required.
-	 */
-	ELF_PLAT_INIT(regs);
-#endif
-
-
-	start_thread32(regs, elf_entry, bprm->p);
-	if (current->flags & PF_PTRACED)
-		send_sig(SIGTRAP, current, 0);
-	return 0;
-}
-
-static int
-load_elf32_binary(struct linux_binprm * bprm, struct pt_regs * regs)
-{
-	int retval;
-
-	MOD_INC_USE_COUNT;
-	retval = do_load_elf32_binary(bprm, regs);
-	MOD_DEC_USE_COUNT;
-	return retval;
-}
-
-/* This is really simpleminded and specialized - we are loading an
-   a.out library that is given an ELF header. */
-
-static inline int
-do_load_elf32_library(int fd){
-	struct file * file;
-	struct elfhdr elf_ex;
-	struct elf_phdr *elf_phdata  =  NULL;
-	struct  inode * inode;
-	unsigned long len;
-	int elf_bss;
-	int retval;
-	unsigned long bss;
-	int error;
-	int i,j, k;
-
-	len = 0;
-	file = current->files->fd[fd];
-	inode = file->f_inode;
-	elf_bss = 0;
-
-	if (!file || !file->f_op)
-		return -EACCES;
-
-	/* seek to the beginning of the file */
-	if (file->f_op->llseek) {
-		if ((error = file->f_op->llseek(inode, file, 0, 0)) != 0)
-			return -ENOEXEC;
-	} else
-		file->f_pos = 0;
-
-	set_fs(KERNEL_DS);
-	error = file->f_op->read(inode, file, (char *) &elf_ex, sizeof(elf_ex));
-	set_fs(USER_DS);
-	if (error != sizeof(elf_ex))
-		return -ENOEXEC;
-
-	if (elf_ex.e_ident[0] != 0x7f ||
-	    strncmp(&elf_ex.e_ident[1], "ELF",3) != 0)
-		return -ENOEXEC;
-
-	/* First of all, some simple consistency checks */
-	if (elf_ex.e_type != ET_EXEC || elf_ex.e_phnum > 2 ||
-	   !elf_check_arch(elf_ex.e_machine) ||
-	   (!inode->i_op || !inode->i_op->default_file_ops->mmap))
-		return -ENOEXEC;
-
-	/* Now read in all of the header information */
-
-	if (sizeof(struct elf_phdr) * elf_ex.e_phnum > PAGE_SIZE)
-		return -ENOEXEC;
-
-	elf_phdata =  (struct elf_phdr *)
-		kmalloc(sizeof(struct elf_phdr) * elf_ex.e_phnum, GFP_KERNEL);
-	if (elf_phdata == NULL)
-		return -ENOMEM;
-
-	retval = read_exec(inode, elf_ex.e_phoff, (char *) elf_phdata,
-			   sizeof(struct elf_phdr) * elf_ex.e_phnum, 1);
-
-	j = 0;
-	for(i=0; i<elf_ex.e_phnum; i++)
-		if ((elf_phdata + i)->p_type == PT_LOAD) j++;
-
-	if (j != 1)  {
-		kfree(elf_phdata);
-		return -ENOEXEC;
-	}
-
-	while(elf_phdata->p_type != PT_LOAD) elf_phdata++;
-
-	/* Now use mmap to map the library into memory. */
-	error = do_mmap(file,
-			ELF_PAGESTART(elf_phdata->p_vaddr),
-			(elf_phdata->p_filesz +
-			 ELF_PAGEOFFSET(elf_phdata->p_vaddr)),
-			PROT_READ | PROT_WRITE | PROT_EXEC,
-			MAP_FIXED | MAP_PRIVATE | MAP_DENYWRITE,
-			(elf_phdata->p_offset -
-			 ELF_PAGEOFFSET(elf_phdata->p_vaddr)));
-
-	k = elf_phdata->p_vaddr + elf_phdata->p_filesz;
-	if (k > elf_bss) elf_bss = k;
-
-	if (error != ELF_PAGESTART(elf_phdata->p_vaddr)) {
-		kfree(elf_phdata);
-		return error;
-	}
-
-	padzero(elf_bss);
-
-	len = ELF_PAGESTART(elf_phdata->p_filesz + elf_phdata->p_vaddr+ ELF_EXEC_PAGESIZE - 1);
-	bss = elf_phdata->p_memsz + elf_phdata->p_vaddr;
-	if (bss > len)
-	  do_mmap(NULL, len, bss-len,
-		  PROT_READ|PROT_WRITE|PROT_EXEC,
-		  MAP_FIXED|MAP_PRIVATE, 0);
-	kfree(elf_phdata);
-	return 0;
-}
-
-static int load_elf32_library(int fd)
-{
-	int retval;
-
-	MOD_INC_USE_COUNT;
-	retval = do_load_elf32_library(fd);
-	MOD_DEC_USE_COUNT;
-	return retval;
-}
-
-/*
- * Note that some platforms still use traditional core dumps and not
- * the ELF core dump.  Each platform can select it as appropriate.
- */
-#ifdef USE_ELF_CORE_DUMP
-
-/*
- * ELF core dumper
- *
- * Modelled on fs/exec.c:aout_core_dump()
- * Jeremy Fitzhardinge <jeremy@sw.oz.au>
- */
-/*
- * These are the only things you should do on a core-file: use only these
- * functions to write out all the necessary info.
- */
-static int dump_write(struct file *file, const void *addr, int nr)
-{
-	return file->f_op->write(file->f_inode, file, addr, nr) == nr;
-}
-
-static int dump_seek(struct file *file, off_t off)
-{
-	if (file->f_op->llseek) {
-		if (file->f_op->llseek(file->f_inode, file, off, 0) != off)
-			return 0;
-	} else
-		file->f_pos = off;
-	return 1;
-}
-
-/*
- * Decide whether a segment is worth dumping; default is yes to be
- * sure (missing info is worse than too much; etc).
- * Personally I'd include everything, and use the coredump limit...
- *
- * I think we should skip something. But I am not sure how. H.J.
- */
-static inline int maydump(struct vm_area_struct *vma)
-{
-	if (!(vma->vm_flags & (VM_READ|VM_WRITE|VM_EXEC)))
-		return 0;
-#if 1
-	if (vma->vm_flags & (VM_WRITE|VM_GROWSUP|VM_GROWSDOWN))
-		return 1;
-	if (vma->vm_flags & (VM_READ|VM_EXEC|VM_EXECUTABLE|VM_SHARED))
-		return 0;
-#endif
-	return 1;
-}
-
-#define roundup(x, y)  ((((x)+((y)-1))/(y))*(y))
-
-/* An ELF note in memory */
-struct memelfnote
-{
-	const char *name;
-	int type;
-	unsigned int datasz;
-	void *data;
-};
-
-static int notesize(struct memelfnote *en)
-{
-	int sz;
-
-	sz = sizeof(struct elf_note);
-	sz += roundup(strlen(en->name), 4);
-	sz += roundup(en->datasz, 4);
-
-	return sz;
-}
-
-/* #define DEBUG */
-
-#define DUMP_WRITE(addr, nr)	\
-	do { if (!dump_write(file, (addr), (nr))) return 0; } while(0)
-#define DUMP_SEEK(off)	\
-	do { if (!dump_seek(file, (off))) return 0; } while(0)
-
-static int writenote(struct memelfnote *men, struct file *file)
-{
-	struct elf_note en;
-
-	en.n_namesz = strlen(men->name);
-	en.n_descsz = men->datasz;
-	en.n_type = men->type;
-
-	DUMP_WRITE(&en, sizeof(en));
-	DUMP_WRITE(men->name, en.n_namesz);
-	/* XXX - cast from long long to long to avoid need for libgcc.a */
-	DUMP_SEEK(roundup((unsigned long)file->f_pos, 4));	/* XXX */
-	DUMP_WRITE(men->data, men->datasz);
-	DUMP_SEEK(roundup((unsigned long)file->f_pos, 4));	/* XXX */
-
-	return 1;
-}
-#undef DUMP_WRITE
-#undef DUMP_SEEK
-
-#define DUMP_WRITE(addr, nr)	\
-	if (!dump_write(&file, (addr), (nr))) \
-		goto close_coredump;
-#define DUMP_SEEK(off)	\
-	if (!dump_seek(&file, (off))) \
-		goto close_coredump;
-/*
- * Actual dumper
- *
- * This is a two-pass process; first we find the offsets of the bits,
- * and then they are actually written out.  If we run out of core limit
- * we just truncate.
- */
-static int elf32_core_dump(long signr, struct pt_regs * regs)
-{
-	int has_dumped = 0;
-	struct file file;
-	struct inode *inode;
-	unsigned short fs;
-	char corefile[6+sizeof(current->comm)];
-	int segs;
-	int i;
-	size_t size;
-	struct vm_area_struct *vma;
-	struct elfhdr elf;
-	off_t offset = 0, dataoff;
-	int limit = current->rlim[RLIMIT_CORE].rlim_cur;
-	int numnote = 4;
-	struct memelfnote notes[4];
-	struct elf_prstatus prstatus;	/* NT_PRSTATUS */
-	elf_fpregset_t fpu;		/* NT_PRFPREG */
-	struct elf_prpsinfo psinfo;	/* NT_PRPSINFO */
-
-	if (!current->dumpable || limit < PAGE_SIZE || current->mm->count != 1)
-		return 0;
-	current->dumpable = 0;
-
-#ifndef CONFIG_BINFMT_ELF32
-	MOD_INC_USE_COUNT;
-#endif
-
-	/* Count what's needed to dump, up to the limit of coredump size */
-	segs = 0;
-	size = 0;
-	for(vma = current->mm->mmap; vma != NULL; vma = vma->vm_next) {
-		if (maydump(vma))
-		{
-			int sz = vma->vm_end-vma->vm_start;
-
-			if (size+sz >= limit)
-				break;
-			else
-				size += sz;
-		}
-
-		segs++;
-	}
-#ifdef DEBUG
-	printk("elf_core_dump: %d segs taking %d bytes\n", segs, size);
-#endif
-
-	/* Set up header */
-	memcpy(elf.e_ident, ELFMAG, SELFMAG);
-	elf.e_ident[EI_CLASS] = ELF_CLASS;
-	elf.e_ident[EI_DATA] = ELF_DATA;
-	elf.e_ident[EI_VERSION] = EV_CURRENT;
-	memset(elf.e_ident+EI_PAD, 0, EI_NIDENT-EI_PAD);
-
-	elf.e_type = ET_CORE;
-	elf.e_machine = ELF_ARCH;
-	elf.e_version = EV_CURRENT;
-	elf.e_entry = 0;
-	elf.e_phoff = sizeof(elf);
-	elf.e_shoff = 0;
-	elf.e_flags = 0;
-	elf.e_ehsize = sizeof(elf);
-	elf.e_phentsize = sizeof(struct elf_phdr);
-	elf.e_phnum = segs+1;		/* Include notes */
-	elf.e_shentsize = 0;
-	elf.e_shnum = 0;
-	elf.e_shstrndx = 0;
-
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	memcpy(corefile,"core.",5);
-#if 0
-	memcpy(corefile+5,current->comm,sizeof(current->comm));
-#else
-	corefile[4] = '\0';
-#endif
-	if (open_namei(corefile,O_CREAT | 2 | O_TRUNC,0600,&inode,NULL)) {
-		inode = NULL;
-		goto end_coredump;
-	}
-	if (!S_ISREG(inode->i_mode))
-		goto end_coredump;
-	if (!inode->i_op || !inode->i_op->default_file_ops)
-		goto end_coredump;
-	file.f_mode = 3;
-	file.f_flags = 0;
-	file.f_count = 1;
-	file.f_inode = inode;
-	file.f_pos = 0;
-	file.f_reada = 0;
-	file.f_op = inode->i_op->default_file_ops;
-	if (file.f_op->open)
-		if (file.f_op->open(inode,&file))
-			goto end_coredump;
-	if (!file.f_op->write)
-		goto close_coredump;
-	has_dumped = 1;
-	current->flags |= PF_DUMPCORE;
-
-	DUMP_WRITE(&elf, sizeof(elf));
-	offset += sizeof(elf);				/* Elf header */
-	offset += (segs+1) * sizeof(struct elf_phdr);	/* Program headers */
-
-	/*
-	 * Set up the notes in similar form to SVR4 core dumps made
-	 * with info from their /proc.
-	 */
-	memset(&psinfo, 0, sizeof(psinfo));
-	memset(&prstatus, 0, sizeof(prstatus));
-
-	notes[0].name = "CORE";
-	notes[0].type = NT_PRSTATUS;
-	notes[0].datasz = sizeof(prstatus);
-	notes[0].data = &prstatus;
-	prstatus.pr_info.si_signo = prstatus.pr_cursig = signr;
-	prstatus.pr_sigpend = current->signal;
-	prstatus.pr_sighold = current->blocked;
-	psinfo.pr_pid = prstatus.pr_pid = current->pid;
-	psinfo.pr_ppid = prstatus.pr_ppid = current->p_pptr->pid;
-	psinfo.pr_pgrp = prstatus.pr_pgrp = current->pgrp;
-	psinfo.pr_sid = prstatus.pr_sid = current->session;
-	prstatus.pr_utime.tv_sec = CT_TO_SECS(current->utime);
-	prstatus.pr_utime.tv_usec = CT_TO_USECS(current->utime);
-	prstatus.pr_stime.tv_sec = CT_TO_SECS(current->stime);
-	prstatus.pr_stime.tv_usec = CT_TO_USECS(current->stime);
-	prstatus.pr_cutime.tv_sec = CT_TO_SECS(current->cutime);
-	prstatus.pr_cutime.tv_usec = CT_TO_USECS(current->cutime);
-	prstatus.pr_cstime.tv_sec = CT_TO_SECS(current->cstime);
-	prstatus.pr_cstime.tv_usec = CT_TO_USECS(current->cstime);
-
-	/*
-	 * This transfers the registers from regs into the standard
-	 * coredump arrangement, whatever that is.
-	 */
-#ifdef ELF_CORE_COPY_REGS
-	ELF_CORE_COPY_REGS(prstatus.pr_reg, regs)
-#else
-	if (sizeof(elf_gregset_t) != sizeof(struct pt_regs))
-	{
-		printk("sizeof(elf_gregset_t) (%ld) != sizeof(struct pt_regs) (%ld)\n",
-			sizeof(elf_gregset_t), sizeof(struct pt_regs));
-	}
-	else
-		*(struct pt_regs *)&prstatus.pr_reg = *regs;
-#endif
-
-	notes[1].name = "CORE";
-	notes[1].type = NT_PRPSINFO;
-	notes[1].datasz = sizeof(psinfo);
-	notes[1].data = &psinfo;
-	psinfo.pr_state = current->state;
-	psinfo.pr_sname = (current->state < 0 || current->state > 5) ? '.' : "RSDZTD"[current->state];
-	psinfo.pr_zomb = psinfo.pr_sname == 'Z';
-	psinfo.pr_nice = current->priority-15;
-	psinfo.pr_flag = current->flags;
-	psinfo.pr_uid = current->uid;
-	psinfo.pr_gid = current->gid;
-	{
-		int i, len;
-
-		set_fs(fs);
-
-		len = current->mm->arg_end - current->mm->arg_start;
-		len = len >= ELF_PRARGSZ ? ELF_PRARGSZ : len;
-		copy_from_user(&psinfo.pr_psargs,
-			      (const char *)current->mm->arg_start, len);
-		for(i = 0; i < len; i++)
-			if (psinfo.pr_psargs[i] == 0)
-				psinfo.pr_psargs[i] = ' ';
-		psinfo.pr_psargs[len] = 0;
-
-		set_fs(KERNEL_DS);
-	}
-	strncpy(psinfo.pr_fname, current->comm, sizeof(psinfo.pr_fname));
-
-	notes[2].name = "CORE";
-	notes[2].type = NT_TASKSTRUCT;
-	notes[2].datasz = sizeof(*current);
-	notes[2].data = current;
-
-	/* Try to dump the fpu. */
-	prstatus.pr_fpvalid = dump_fpu (regs, &fpu);
-	if (!prstatus.pr_fpvalid)
-	{
-		numnote--;
-	}
-	else
-	{
-		notes[3].name = "CORE";
-		notes[3].type = NT_PRFPREG;
-		notes[3].datasz = sizeof(fpu);
-		notes[3].data = &fpu;
-	}
-	
-	/* Write notes phdr entry */
-	{
-		struct elf_phdr phdr;
-		int sz = 0;
-
-		for(i = 0; i < numnote; i++)
-			sz += notesize(&notes[i]);
-
-		phdr.p_type = PT_NOTE;
-		phdr.p_offset = offset;
-		phdr.p_vaddr = 0;
-		phdr.p_paddr = 0;
-		phdr.p_filesz = sz;
-		phdr.p_memsz = 0;
-		phdr.p_flags = 0;
-		phdr.p_align = 0;
-
-		offset += phdr.p_filesz;
-		DUMP_WRITE(&phdr, sizeof(phdr));
-	}
-
-	/* Page-align dumped data */
-	dataoff = offset = roundup(offset, PAGE_SIZE);
-
-	/* Write program headers for segments dump */
-	for(vma = current->mm->mmap, i = 0;
-		i < segs && vma != NULL; vma = vma->vm_next) {
-		struct elf_phdr phdr;
-		size_t sz;
-
-		i++;
-
-		sz = vma->vm_end - vma->vm_start;
-
-		phdr.p_type = PT_LOAD;
-		phdr.p_offset = offset;
-		phdr.p_vaddr = vma->vm_start;
-		phdr.p_paddr = 0;
-		phdr.p_filesz = maydump(vma) ? sz : 0;
-		phdr.p_memsz = sz;
-		offset += phdr.p_filesz;
-		phdr.p_flags = vma->vm_flags & VM_READ ? PF_R : 0;
-		if (vma->vm_flags & VM_WRITE) phdr.p_flags |= PF_W;
-		if (vma->vm_flags & VM_EXEC) phdr.p_flags |= PF_X;
-		phdr.p_align = PAGE_SIZE;
-
-		DUMP_WRITE(&phdr, sizeof(phdr));
-	}
-
-	for(i = 0; i < numnote; i++)
-		if (!writenote(&notes[i], &file))
-			goto close_coredump;
-
-	set_fs(fs);
-
-	DUMP_SEEK(dataoff);
-
-	for(i = 0, vma = current->mm->mmap;
-	    i < segs && vma != NULL;
-	    vma = vma->vm_next) {
-		unsigned long addr = vma->vm_start;
-		unsigned long len = vma->vm_end - vma->vm_start;
-
-		i++;
-		if (!maydump(vma))
-			continue;
-#ifdef DEBUG
-		printk("elf_core_dump: writing %08lx %lx\n", addr, len);
-#endif
-		DUMP_WRITE((void *)addr, len);
-	}
-
-	if ((off_t) file.f_pos != offset) {
-		/* Sanity check */
-		printk("elf_core_dump: file.f_pos (%ld) != offset (%ld)\n",
-		       (off_t) file.f_pos, offset);
-	}
-
- close_coredump:
-	if (file.f_op->release)
-		file.f_op->release(inode,&file);
-
- end_coredump:
-	set_fs(fs);
-	iput(inode);
-#ifndef CONFIG_BINFMT_ELF32
-	MOD_DEC_USE_COUNT;
-#endif
-	return has_dumped;
-}
-#endif		/* USE_ELF_CORE_DUMP */
-
-__initfunc(int init_elf32_binfmt(void))
-{
-	return register_binfmt(&elf32_format);
-}
-
-#ifdef MODULE
-
-int init_module(void)
-{
-	/* Install the COFF, ELF and XOUT loaders.
-	 * N.B. We *rely* on the table being the right size with the
-	 * right number of free slots...
-	 */
-	return init_elf32_binfmt();
-}
-
-
-void cleanup_module( void)
-{
-	/* Remove the COFF and ELF loaders. */
-	unregister_binfmt(&elf32_format);
-}
-#endif
diff --git a/arch/sparc64/kernel/sys_sparc32.c b/arch/sparc64/kernel/sys_sparc32.c
index e9911daed..147b60c34 100644
--- a/arch/sparc64/kernel/sys_sparc32.c
+++ b/arch/sparc64/kernel/sys_sparc32.c
@@ -1,4 +1,4 @@
-/* $Id: sys_sparc32.c,v 1.9 1997/04/21 08:34:24 jj Exp $
+/* $Id: sys_sparc32.c,v 1.13 1997/05/18 04:16:44 davem Exp $
  * sys_sparc32.c: Conversion between 32bit and 64bit native syscalls.
  *
  * Copyright (C) 1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
@@ -27,15 +27,19 @@
 #include <linux/nfs_fs.h>
 #include <linux/smb_fs.h>
 #include <linux/ncp_fs.h>
+#include <linux/quota.h>
 
 #include <asm/types.h>
 #include <asm/poll.h>
 #include <asm/ipc.h>
 #include <asm/uaccess.h>
 
-/* As gcc will warn about casting u32 to some ptr, we have to cast it to unsigned long first, and that's what is A() for.
- * You just do (void *)A(x), instead of having to type (void *)((unsigned long)x) or instead of just (void *)x, which will
- * produce warnings */
+/* As gcc will warn about casting u32 to some ptr, we have to cast it to
+ * unsigned long first, and that's what is A() for.
+ * You just do (void *)A(x), instead of having to
+ * type (void *)((unsigned long)x) or instead of just (void *)x, which will
+ * produce warnings.
+ */
 #define A(x) ((unsigned long)x)
  
 extern asmlinkage int sys_ioperm(unsigned long from, unsigned long num, int on);
@@ -138,6 +142,10 @@ extern asmlinkage int sys_sendmsg(int fd, struct msghdr *msg, unsigned flags);
 extern asmlinkage int sys_recvmsg(int fd, struct msghdr *msg, unsigned int flags);
 extern asmlinkage int sys_socketcall(int call, unsigned long *args);
 extern asmlinkage int sys_nfsservctl(int cmd, void *argp, void *resp);
+extern asmlinkage int sys_listen(int fd, int backlog);
+extern asmlinkage int sys_socket(int family, int type, int protocol);
+extern asmlinkage int sys_socketpair(int family, int type, int protocol, int usockvec[2]);
+extern asmlinkage int sys_shutdown(int fd, int how);
 
 asmlinkage int sys32_ioperm(u32 from, u32 num, int on)
 {
@@ -157,6 +165,17 @@ struct ipc_perm32
         unsigned short  seq;
 };
 
+struct semid_ds32 {
+        struct ipc_perm32 sem_perm;               /* permissions .. see ipc.h */
+        __kernel_time_t32 sem_otime;              /* last semop time */
+        __kernel_time_t32 sem_ctime;              /* last change time */
+        u32 sem_base;              /* ptr to first semaphore in array */
+        u32 sem_pending;          /* pending operations to be processed */
+        u32 sem_pending_last;    /* last pending operation */
+        u32 undo;                  /* undo requests on this array */
+        unsigned short  sem_nsems;              /* no. of semaphores in array */
+};
+
 struct msqid_ds32
 {
         struct ipc_perm32 msg_perm;
@@ -212,15 +231,62 @@ asmlinkage int sys32_ipc (u32 call, int first, int second, int third, u32 ptr, u
 			err = sys_semget (first, second, third);
 			goto out;
 		case SEMCTL: {
-			/* XXX union semun32 to union semun64 and back conversion */
 			union semun fourth;
+			void *pad;
+			unsigned long old_fs;
+			struct semid_ds s;
+			
 			err = -EINVAL;
 			if (!ptr)
 				goto out;
 			err = -EFAULT;
-			if(get_user(fourth.__pad, (void **)A(ptr)))
+			if(get_user(pad, (void **)A(ptr)))
 				goto out;
+			fourth.__pad = pad;
+			switch (third) {
+				case IPC_INFO:
+				case SEM_INFO:
+				case GETVAL:
+				case GETPID:
+				case GETNCNT:
+				case GETZCNT:
+				case GETALL:
+				case SETALL:
+				case IPC_RMID:
+					err = sys_semctl (first, second, third, fourth);
+					goto out;
+				case IPC_SET:
+					if (get_user (s.sem_perm.uid, &(((struct semid_ds32 *)A(pad))->sem_perm.uid)) ||
+					    __get_user (s.sem_perm.gid, &(((struct semid_ds32 *)A(pad))->sem_perm.gid)) ||
+					    __get_user (s.sem_perm.mode, &(((struct semid_ds32 *)A(pad))->sem_perm.mode))) {
+						err = -EFAULT;
+						goto out;
+					}
+					/* Fall through */
+				case SEM_STAT:
+				case IPC_STAT:
+					fourth.__pad = &s;
+					break;
+			}
+			old_fs = get_fs();
+			set_fs (KERNEL_DS);
 			err = sys_semctl (first, second, third, fourth);
+			set_fs (old_fs);
+			switch (third) {
+				case SEM_STAT:
+				case IPC_STAT:
+					if (put_user (s.sem_perm.key, &(((struct semid_ds32 *)A(pad))->sem_perm.key)) ||
+					    __put_user (s.sem_perm.uid, &(((struct semid_ds32 *)A(pad))->sem_perm.uid)) ||
+					    __put_user (s.sem_perm.gid, &(((struct semid_ds32 *)A(pad))->sem_perm.gid)) ||
+					    __put_user (s.sem_perm.cuid, &(((struct semid_ds32 *)A(pad))->sem_perm.cuid)) ||
+					    __put_user (s.sem_perm.cgid, &(((struct semid_ds32 *)A(pad))->sem_perm.cgid)) ||
+					    __put_user (s.sem_perm.mode, &(((struct semid_ds32 *)A(pad))->sem_perm.mode)) ||
+					    __put_user (s.sem_perm.seq, &(((struct semid_ds32 *)A(pad))->sem_perm.seq)) ||
+					    __put_user (s.sem_otime, &(((struct semid_ds32 *)A(pad))->sem_otime)) ||
+					    __put_user (s.sem_ctime, &(((struct semid_ds32 *)A(pad))->sem_ctime)) ||
+					    __put_user (s.sem_nsems, &(((struct semid_ds32 *)A(pad))->sem_nsems)))
+						err = -EFAULT;
+			}
 			goto out;
 			}
 		default:
@@ -534,10 +600,50 @@ asmlinkage int sys32_rename(u32 oldname, u32 newname)
 	return sys_rename((const char *)A(oldname), (const char *)A(newname));
 }
 
-/* XXX: Play with the addr, it will be ugly :(( */
+struct dqblk32 {
+    __u32 dqb_bhardlimit;
+    __u32 dqb_bsoftlimit;
+    __u32 dqb_curblocks;
+    __u32 dqb_ihardlimit;
+    __u32 dqb_isoftlimit;
+    __u32 dqb_curinodes;
+    __kernel_time_t32 dqb_btime;
+    __kernel_time_t32 dqb_itime;
+};
+                                
 asmlinkage int sys32_quotactl(int cmd, u32 special, int id, u32 addr)
 {
-	return sys_quotactl(cmd, (const char *)A(special), id, (caddr_t)A(addr));
+	int cmds = cmd >> SUBCMDSHIFT;
+	int err;
+	struct dqblk d;
+	unsigned long old_fs;
+	
+	switch (cmds) {
+	case Q_GETQUOTA:
+		break;
+	case Q_SETQUOTA:
+	case Q_SETUSE:
+	case Q_SETQLIM:
+		if (copy_from_user (&d, (struct dqblk32 *)A(addr), sizeof (struct dqblk32)))
+			return -EFAULT;
+		d.dqb_itime = ((struct dqblk32 *)&d)->dqb_itime;
+		d.dqb_btime = ((struct dqblk32 *)&d)->dqb_btime;
+		break;
+	default:
+		return sys_quotactl(cmd, (const char *)A(special), id, (caddr_t)A(addr));
+	}
+	old_fs = get_fs ();
+	set_fs (KERNEL_DS);
+	err = sys_quotactl(cmd, (const char *)A(special), id, (caddr_t)A(addr));
+	set_fs (old_fs);
+	if (cmds == Q_GETQUOTA) {
+		__kernel_time_t b = d.dqb_btime, i = d.dqb_itime;
+		((struct dqblk32 *)&d)->dqb_itime = i;
+		((struct dqblk32 *)&d)->dqb_btime = b;
+		if (copy_to_user ((struct dqblk32 *)A(addr), &d, sizeof (struct dqblk32)))
+			return -EFAULT;
+	}
+	return err;
 }
 
 static int put_statfs (u32 buf, struct statfs *s)
@@ -1599,20 +1705,224 @@ asmlinkage int sys32_getsockopt(int fd, int level, int optname, u32 optval, u32
 	return sys_getsockopt(fd, level, optname, (char *)A(optval), (int *)A(optlen));
 }
 
-/* Continue here */
+struct msghdr32 {
+        u32               msg_name;
+        int               msg_namelen;
+        u32               msg_iov;
+        __kernel_size_t32 msg_iovlen;
+        u32               msg_control;
+        __kernel_size_t32 msg_controllen;
+        unsigned          msg_flags;
+};
+
+struct cmsghdr32 {
+        __kernel_size_t32 cmsg_len;
+        int               cmsg_level;
+        int               cmsg_type;
+        unsigned char     cmsg_data[0];
+};
+
 asmlinkage int sys32_sendmsg(int fd, u32 msg, unsigned flags)
 {
-	return sys_sendmsg(fd, (struct msghdr *)A(msg), flags);
+	struct msghdr m;
+	int count;
+	struct iovec *v;
+	struct iovec vf[UIO_FASTIOV];
+	u32 i, vector;
+	long ret;
+	unsigned long old_fs;
+	
+	if (get_user ((long)m.msg_name, &(((struct msghdr32 *)A(msg))->msg_name)) ||
+	    __get_user (m.msg_namelen, &(((struct msghdr32 *)A(msg))->msg_namelen)) ||
+	    __get_user (vector, &(((struct msghdr32 *)A(msg))->msg_iov)) ||
+	    __get_user (m.msg_iovlen, &(((struct msghdr32 *)A(msg))->msg_iovlen)) ||
+	    __get_user ((long)m.msg_control, &(((struct msghdr32 *)A(msg))->msg_control)) ||
+	    __get_user (m.msg_controllen, &(((struct msghdr32 *)A(msg))->msg_controllen)) ||
+	    __get_user (m.msg_flags, &(((struct msghdr32 *)A(msg))->msg_flags)))
+		return -EFAULT;
+	
+	count = m.msg_iovlen;
+	if (!count) return 0; if (count > UIO_MAXIOV) return -EINVAL;
+	if (count <= UIO_FASTIOV)
+		v = vf;
+	else {
+		lock_kernel ();
+		v = kmalloc (count * sizeof (struct iovec), GFP_KERNEL);
+		if (!v) {
+			ret = -ENOMEM;
+			goto out;
+		}
+	}
+
+	for (i = 0; i < count; i++) {
+		if (__get_user ((unsigned long)(v[i].iov_base), &((((struct iovec32 *)A(vector))+i)->iov_base)) ||
+		    __get_user (v[i].iov_len, &((((struct iovec32 *)A(vector))+i)->iov_len))) {
+		    	ret = -EFAULT;
+		    	goto out;
+		}
+	}
+	
+	m.msg_iov = v;
+
+	if (m.msg_controllen) {
+		/* XXX Handle msg_control stuff... Or should that go into ip_sockglue.c etc.? */
+	}
+	old_fs = get_fs();
+	set_fs (KERNEL_DS);
+	ret = sys_sendmsg(fd, &m, flags);
+	set_fs (old_fs);
+out:
+	if (count > UIO_FASTIOV) {
+		kfree (v);
+		unlock_kernel ();
+	}
+	return ret;
 }
 
 asmlinkage int sys32_recvmsg(int fd, u32 msg, unsigned int flags)
 {
-	return sys_recvmsg(fd, (struct msghdr *)A(msg), flags);
+	struct msghdr m;
+	int count;
+	struct iovec *v;
+	struct iovec vf[UIO_FASTIOV];
+	u32 i, vector;
+	long ret;
+	unsigned long old_fs;
+	
+	if (get_user ((long)m.msg_name, &(((struct msghdr32 *)A(msg))->msg_name)) ||
+	    __get_user (m.msg_namelen, &(((struct msghdr32 *)A(msg))->msg_namelen)) ||
+	    __get_user (vector, &(((struct msghdr32 *)A(msg))->msg_iov)) ||
+	    __get_user (m.msg_iovlen, &(((struct msghdr32 *)A(msg))->msg_iovlen)) ||
+	    __get_user ((long)m.msg_control, &(((struct msghdr32 *)A(msg))->msg_control)) ||
+	    __get_user (m.msg_controllen, &(((struct msghdr32 *)A(msg))->msg_controllen)) ||
+	    __get_user (m.msg_flags, &(((struct msghdr32 *)A(msg))->msg_flags)))
+		return -EFAULT;
+	
+	count = m.msg_iovlen;
+	if (!count) return 0; if (count > UIO_MAXIOV) return -EINVAL;
+	if (count <= UIO_FASTIOV)
+		v = vf;
+	else {
+		lock_kernel ();
+		v = kmalloc (count * sizeof (struct iovec), GFP_KERNEL);
+		if (!v) {
+			ret = -ENOMEM;
+			goto out;
+		}
+	}
+
+	for (i = 0; i < count; i++) {
+		if (__get_user ((unsigned long)(v[i].iov_base), &((((struct iovec32 *)A(vector))+i)->iov_base)) ||
+		    __get_user (v[i].iov_len, &((((struct iovec32 *)A(vector))+i)->iov_len))) {
+		    	ret = -EFAULT;
+		    	goto out;
+		}
+	}
+	
+	m.msg_iov = v;
+
+	if (m.msg_controllen) {
+		/* XXX Handle msg_control stuff... Or should that go into ip_sockglue.c etc.? */
+	}
+	old_fs = get_fs();
+	set_fs (KERNEL_DS);
+	ret = sys_recvmsg(fd, &m, flags);
+	set_fs (old_fs);
+	if (ret >= 0) {
+		/* XXX Handle msg_control stuff... */
+		if (put_user (m.msg_flags, &(((struct msghdr32 *)A(msg))->msg_flags)) ||
+		    __put_user (m.msg_controllen, &(((struct msghdr32 *)A(msg))->msg_controllen)))
+			return -EFAULT;
+	}
+out:
+	if (count > UIO_FASTIOV) {
+		kfree (v);
+		unlock_kernel ();
+	}
+	return ret;
 }
 
 asmlinkage int sys32_socketcall(int call, u32 args)
 {
-	return sys_socketcall(call, (unsigned long *)A(args));
+	static unsigned char nargs[18]={0,3,3,3,2,3,3,3,
+				        4,4,4,6,6,2,5,5,3,3};
+	u32 a[6];
+	u32 a0,a1;
+	int err = -EINVAL;
+	int i;
+				 
+	lock_kernel();
+	if(call<1||call>SYS_RECVMSG)
+		goto out;
+	err = -EFAULT;
+
+	for (i = 0; i < nargs[call]; i++, args += sizeof (u32))
+		if (get_user(a[i], (u32 *)A(args)))
+			goto out;
+		
+	a0=a[0];
+	a1=a[1];
+	
+	switch(call) 
+	{
+		case SYS_SOCKET:
+			err = sys_socket(a0, a1, a[2]);
+			break;
+		case SYS_BIND:
+			err = sys32_bind(a0, a1, a[2]);
+			break;
+		case SYS_CONNECT:
+			err = sys32_connect(a0, a1, a[2]);
+			break;
+		case SYS_LISTEN:
+			err = sys_listen(a0, a1);
+			break;
+		case SYS_ACCEPT:
+			err = sys32_accept(a0, a1, a[2]);
+			break;
+		case SYS_GETSOCKNAME:
+			err = sys32_getsockname(a0, a1, a[2]);
+			break;
+		case SYS_GETPEERNAME:
+			err = sys32_getpeername(a0, a1, a[2]);
+			break;
+		case SYS_SOCKETPAIR:
+			err = sys_socketpair(a0, a1, a[2], (int *)A(a[3]));
+			break;
+		case SYS_SEND:
+			err = sys32_send(a0, a1, a[2], a[3]);
+			break;
+		case SYS_SENDTO:
+			err = sys32_sendto(a0, a1, a[2], a[3], a[4], a[5]);
+			break;
+		case SYS_RECV:
+			err = sys32_recv(a0, a1, a[2], a[3]);
+			break;
+		case SYS_RECVFROM:
+			err = sys32_recvfrom(a0, a1, a[2], a[3], a[4], a[5]);
+			break;
+		case SYS_SHUTDOWN:
+			err = sys_shutdown(a0,a1);
+			break;
+		case SYS_SETSOCKOPT:
+			err = sys32_setsockopt(a0, a1, a[2], a[3], a[4]);
+			break;
+		case SYS_GETSOCKOPT:
+			err = sys32_getsockopt(a0, a1, a[2], a[3], a[4]);
+			break;
+		case SYS_SENDMSG:
+			err = sys32_sendmsg(a0, a1, a[2]);
+			break;
+		case SYS_RECVMSG:
+			err = sys32_recvmsg(a0, a1, a[2]);
+			break;
+		default:
+			err = -EINVAL;
+			break;
+	}
+out:
+	unlock_kernel();
+	return err;
 }
 
 extern void check_pending(int signum);
diff --git a/arch/sparc64/kernel/traps.c b/arch/sparc64/kernel/traps.c
index 48648c39d..c9774df06 100644
--- a/arch/sparc64/kernel/traps.c
+++ b/arch/sparc64/kernel/traps.c
@@ -1,4 +1,4 @@
-/* $Id: traps.c,v 1.5 1997/04/14 06:56:55 davem Exp $
+/* $Id: traps.c,v 1.10 1997/05/18 08:42:16 davem Exp $
  * arch/sparc/kernel/traps.c
  *
  * Copyright (C) 1995 David S. Miller (davem@caip.rutgers.edu)
@@ -22,6 +22,7 @@
 #include <asm/page.h>
 #include <asm/pgtable.h>
 #include <asm/unistd.h>
+#include <asm/uaccess.h>
 
 /* #define TRAP_DEBUG */
 
@@ -42,6 +43,8 @@ void syscall_trace_entry(struct pt_regs *regs)
 
 void syscall_trace_exit(struct pt_regs *regs)
 {
+	printk("Syscall return check, reg dump.\n");
+	show_regs(regs);
 }
 
 void sparc64_dtlb_fault_handler (void)
@@ -116,25 +119,45 @@ void die_if_kernel(char *str, struct pt_regs *regs)
 	show_regs(regs);
 	printk("Instruction DUMP:");
 	instruction_dump ((unsigned int *) regs->tpc);
+	while(1)
+		barrier();
 	if(regs->tstate & TSTATE_PRIV)
 		do_exit(SIGKILL);
 	do_exit(SIGSEGV);
 }
 
-void do_illegal_instruction(struct pt_regs *regs, unsigned long pc, unsigned long npc,
-			    unsigned long tstate)
+void do_illegal_instruction(struct pt_regs *regs)
 {
+	unsigned long pc = regs->tpc;
+	unsigned long tstate = regs->tstate;
+
 	lock_kernel();
 	if(tstate & TSTATE_PRIV)
 		die_if_kernel("Kernel illegal instruction", regs);
-#ifdef TRAP_DEBUG
-	printk("Ill instr. at pc=%016lx instruction is %08x\n",
-	       regs->tpc, *(unsigned int *)regs->tpc);
+#if 1
+	{
+		unsigned int insn;
+
+		printk("Ill instr. at pc=%016lx ", pc);
+		get_user(insn, ((unsigned int *)pc));
+		printk("insn=[%08x]\n", insn);
+	}
 #endif
 	current->tss.sig_address = pc;
 	current->tss.sig_desc = SUBSIG_ILLINST;
 	send_sig(SIGILL, current, 1);
 	unlock_kernel();
+
+	while(1)
+		barrier();
+}
+
+void do_mna(struct pt_regs *regs)
+{
+	printk("AIEEE: do_mna at %016lx\n", regs->tpc);
+	show_regs(regs);
+	while(1)
+		barrier();
 }
 
 void do_priv_instruction(struct pt_regs *regs, unsigned long pc, unsigned long npc,
diff --git a/arch/sparc64/kernel/ttable.S b/arch/sparc64/kernel/ttable.S
index f22d85014..326382c3f 100644
--- a/arch/sparc64/kernel/ttable.S
+++ b/arch/sparc64/kernel/ttable.S
@@ -1,4 +1,4 @@
-/* $Id: ttable.S,v 1.11 1997/03/25 09:47:21 davem Exp $
+/* $Id: ttable.S,v 1.12 1997/05/17 08:22:30 davem Exp $
  * ttable.S: Sparc V9 Trap Table(s) with SpitFire extensions.
  *
  * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
@@ -13,7 +13,7 @@ tl0_iax:	ACCESS_EXCEPTION_TRAP(instruction_access_exception)
 tl0_resv009:	BTRAP(0x9)
 tl0_iae:	TRAP(do_iae)
 tl0_resv00b:	BTRAP(0xb) BTRAP(0xc) BTRAP(0xd) BTRAP(0xe) BTRAP(0xf)
-tl0_ill:	TRAP(do_ill)
+tl0_ill:	TRAP(do_illegal_instruction)
 tl0_privop:	TRAP(do_privop)
 tl0_resv012:	BTRAP(0x12) BTRAP(0x13) BTRAP(0x14) BTRAP(0x15) BTRAP(0x16) BTRAP(0x17)
 tl0_resv018:	BTRAP(0x18) BTRAP(0x19) BTRAP(0x1a) BTRAP(0x1b) BTRAP(0x1c) BTRAP(0x1d)
@@ -226,27 +226,3 @@ tl1_f4o:	FILL_4_OTHER
 tl1_f5o:	FILL_5_OTHER
 tl1_f6o:	FILL_6_OTHER
 tl1_f7o:	FILL_7_OTHER
-
-#if 0
-/* Unless we are going to have software trap insns in the kernel code, we
- * don't need this. For now we just save 8KB.
- */
-
-#define BTRAPSTL1(x) BTRAPTL1(x) BTRAPTL1(x+1) BTRAPTL1(x+2) BTRAPTL1(x+3) BTRAPTL1(x+4) BTRAPTL1(x+5) BTRAPTL1(x+6) BTRAPTL1(x+7)
-
-tl1_sunos:	BTRAPTL1(0x100)
-tl1_bkpt:	BREAKPOINT_TRAP
-tl1_resv102:	BTRAPTL1(0x102)
-tl1_flushw:	FLUSH_WINDOW_TRAP
-tl1_resv104:	BTRAPTL1(0x104) BTRAPTL1(0x105) BTRAPTL1(0x106)
-tl1_resv107:	BTRAPTL1(0x107) BTRAPTL1(0x108) BTRAPTL1(0x109) BTRAPTL1(0x10a)
-tl1_resv10b:	BTRAPTL1(0x10b) BTRAPTL1(0x10c) BTRAPTL1(0x10d) BTRAPTL1(0x10e)
-tl1_resv10f:	BTRAPTL1(0x10f) 
-tl1_resv110:	BTRAPSTL1(0x110) BTRAPSTL1(0x118)
-tl1_resv120:	BTRAPSTL1(0x120) BTRAPSTL1(0x128)
-tl1_resv130:	BTRAPSTL1(0x130) BTRAPSTL1(0x138)
-tl1_resv140:	BTRAPSTL1(0x140) BTRAPSTL1(0x148)
-tl1_resv150:	BTRAPSTL1(0x150) BTRAPSTL1(0x158)
-tl1_resv160:	BTRAPSTL1(0x160) BTRAPSTL1(0x168)
-tl1_resv170:	BTRAPSTL1(0x170) BTRAPSTL1(0x178)
-#endif
diff --git a/arch/sparc64/kernel/winfixup.S b/arch/sparc64/kernel/winfixup.S
new file mode 100644
index 000000000..a8293c453
--- /dev/null
+++ b/arch/sparc64/kernel/winfixup.S
@@ -0,0 +1,101 @@
+/* $Id: winfixup.S,v 1.3 1997/05/18 22:52:26 davem Exp $
+ *
+ * winfixup.S: Handle cases where user stack pointer is found to be bogus.
+ *
+ * Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
+ */
+
+#include <asm/asi.h>
+#include <asm/head.h>
+#include <asm/page.h>
+#include <asm/ptrace.h>
+#include <asm/processor.h>
+#include <asm/asm_offsets.h>
+
+	.text
+	.align	32
+
+	/* Here are the rules, pay attention.
+	 *
+	 * The kernel is disallowed from touching user space while
+	 * the trap level is greater than zero, except for from within
+	 * the window spill/fill handlers.  This must be followed
+	 * so that we can easily detect the case where we tried to
+	 * spill/fill with a bogus (or unmapped) user stack pointer.
+	 *
+	 * These are layed out in a special way for cache reasons,
+	 * don't touch...
+	 */
+	.globl	winfix_trampoline, fill_fixup, spill_fixup
+fill_fixup:
+	ba,pt	%xcc, etrap
+	 rd	%pc, %g7
+	mov	%l5, %o4
+	mov	%l4, %o5
+	srlx	%l5, PAGE_SHIFT, %o3
+	clr	%o1
+	sllx	%o3, PAGE_SHIFT, %o3
+	and	%l4, 0x4, %o2
+
+	call	do_sparc64_fault
+	 add	%sp, STACK_BIAS + REGWIN_SZ, %o0
+	ba,a,pt	%xcc, rtrap
+	nop
+winfix_trampoline:
+	andn		%g5, 0x7f, %g5
+	add		%g5, 0x7c, %g5
+	wrpr		%g5, %tnpc
+	done
+
+spill_fixup:
+	rd	%pic, %g1
+	ldx	[%g1 + AOFF_task_tss + AOFF_thread_w_saved], %g2
+	sll	%g2, 3, %g5
+	ldx	[%g1 + AOFF_task_tss + AOFF_thread_flags], %g7
+	add	%g1, %g5, %g5
+	andcc	%g7, SPARC_FLAG_32BIT, %g0
+	stx	%sp, [%g5 + AOFF_task_tss + AOFF_thread_rwbuf_stkptrs]
+	sll	%g2, 5, %g5
+
+	bne,pt	%xcc, 1f
+	 add	%g1, %g5, %g5
+	stx	%l0, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x00]
+	stx	%l1, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x08]
+	stx	%l2, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x10]
+	stx	%l3, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x18]
+	stx	%l4, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x20]
+	stx	%l5, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x28]
+
+	stx	%l6, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x30]
+	stx	%l7, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x38]
+	stx	%i0, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x40]
+	stx	%i1, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x48]
+	stx	%i2, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x50]
+	stx	%i3, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x58]
+	stx	%i4, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x60]
+	stx	%i5, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x68]
+
+	stx	%i6, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x70]
+	stx	%i7, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x78]
+	b,a,pt	%xcc, 2f
+	 add	%g2, 1, %g2
+1:
+	std	%l0, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x00]
+	std	%l2, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x08]
+	std	%l4, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x10]
+	std	%l6, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x18]
+
+	std	%i0, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x20]
+	std	%i2, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x28]
+	std	%i4, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x30]
+	std	%i6, [%g5 + AOFF_task_tss + AOFF_thread_reg_window + 0x38]
+	add	%g2, 1, %g2
+2:
+	stx	%g2, [%g1 + AOFF_task_tss + AOFF_thread_w_saved]
+	rdpr	%tstate, %g1
+	nop
+
+	andcc	%g1, TSTATE_PRIV, %g0
+	be,pn	%xcc, fill_fixup
+	 saved
+	retry
diff --git a/arch/sparc64/lib/blockops.S b/arch/sparc64/lib/blockops.S
index b3f06c18d..d0f023d1b 100644
--- a/arch/sparc64/lib/blockops.S
+++ b/arch/sparc64/lib/blockops.S
@@ -1,4 +1,4 @@
-/* $Id: blockops.S,v 1.5 1997/03/26 18:34:28 jj Exp $
+/* $Id: blockops.S,v 1.6 1997/05/18 04:16:49 davem Exp $
  * arch/sparc64/lib/blockops.S: UltraSparc block zero optimized routines.
  *
  * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
@@ -31,37 +31,8 @@
 	.text
 	.align	4
 
-	.globl		bzero_2page, bzero_1page
-bzero_2page:
-	/* %o0 = buf */
-	mov		%o0, %o1
-	wr		%g0, ASI_BLK_P, %asi
-	mov		0x10, %g2
-
-	membar		#Sync|#StoreLoad
-
-	fzero		%f48
-	fzero		%f50
-	fzero		%f52
-	fzero		%f54
-	fzero		%f56
-	fzero		%f58
-	fzero		%f60
-	fzero		%f62
-1:
-	BLAST_BLOCK(%o0, 0x000)
-	BLAST_BLOCK(%o0, 0x100)
-	BLAST_BLOCK(%o0, 0x200)
-	BLAST_BLOCK(%o0, 0x300)
-	subcc		%g2, 1, %g2
-	bne,pt		%icc, 1b
-	 add		%o0, 0x400, %o0
-
-	membar		#Sync|#LoadStore|#StoreStore
-
-	retl
-	 mov		%o1, %o0
-
+#if 0
+	.globl		bzero_1page
 bzero_1page:
 	/* %o0 = buf */
 	mov		%o0, %o1
@@ -89,9 +60,36 @@ bzero_1page:
 
 	retl
 	 mov		%o1, %o0
+#endif
 
 	.globl		__bfill64
 __bfill64:
+#if 1
+	/* %o0 = buf, %o1 = 64-bit pattern */
+#define FILL_BLOCK(buf, offset) \
+	stx	%o1, [buf + offset + 0x38]; \
+	stx	%o1, [buf + offset + 0x30]; \
+	stx	%o1, [buf + offset + 0x28]; \
+	stx	%o1, [buf + offset + 0x20]; \
+	stx	%o1, [buf + offset + 0x18]; \
+	stx	%o1, [buf + offset + 0x10]; \
+	stx	%o1, [buf + offset + 0x08]; \
+	stx	%o1, [buf + offset + 0x00];
+
+	mov	0x20, %g2
+1:
+	FILL_BLOCK(%o0, 0x00)
+	FILL_BLOCK(%o0, 0x40)
+	FILL_BLOCK(%o0, 0x80)
+	FILL_BLOCK(%o0, 0xc0)
+	subcc	%g2, 1, %g2
+	bne,pt	%icc, 1b
+	 add	%o0, 0x100, %o0
+	retl
+	 nop
+#undef FILL_BLOCK
+
+#else
 	/* %o0 = buf */
 	stx		%o1, [%sp + 0x7ff + 128]
 	wr		%g0, ASI_BLK_P, %asi
@@ -116,7 +114,9 @@ __bfill64:
 
 	retl
 	 membar		#Sync|#LoadStore|#StoreStore
+#endif
 
+#if 0
 	.globl		__copy_1page
 __copy_1page:
 	/* %o0 = dst, %o1 = src */
@@ -135,4 +135,4 @@ __copy_1page:
 
 	retl
 	 membar		#Sync|#LoadStore|#StoreStore
-
+#endif
diff --git a/arch/sparc64/lib/checksum.S b/arch/sparc64/lib/checksum.S
index 8a06003ee..b63f0d6e8 100644
--- a/arch/sparc64/lib/checksum.S
+++ b/arch/sparc64/lib/checksum.S
@@ -44,13 +44,13 @@
 csum_partial_end_cruft:
 	andcc	%o1, 8, %g0			! check how much
 	be,pn	%icc, 1f			! caller asks %o1 & 0x8
-	 and	%o1, 4, %g3			! nope, check for word remaining
+	 and	%o1, 4, %g5			! nope, check for word remaining
 	ldd	[%o0], %g2			! load two
 	addcc	%g2, %o2, %o2			! add first word to sum
 	addccc	%g3, %o2, %o2			! add second word as well
 	add	%o0, 8, %o0			! advance buf ptr
 	addc	%g0, %o2, %o2			! add in final carry
-1:	brz,pn	%g3, 1f				! nope, skip this code
+1:	brz,pn	%g5, 1f				! nope, skip this code
 	 andcc	%o1, 3, %o1			! check for trailing bytes
 	ld	[%o0], %g2			! load it
 	addcc	%g2, %o2, %o2			! add to sum
@@ -98,15 +98,17 @@ csum_partial:						/* %o0=buf, %o1=len, %o2=sum */
 	srl	%o2, 16, %g3
 	addc	%g0, %g3, %g2
 	sll	%o2, 16, %o2
+	and	%o0, 0x4, %g7
 	sll	%g2, 16, %g3
 	srl	%o2, 16, %o2
 	or	%g3, %o2, %o2
 1:	brz,pn	%g7, csum_partial_fix_aligned
-	 nop
+	 andn	%o1, 0x7f, %o3
 	ld	[%o0 + 0x00], %g2
 	sub	%o1, 4, %o1
 	addcc	%g2, %o2, %o2
 	add	%o0, 4, %o0
+	andn	%o1, 0x7f, %o3
 	addc	%g0, %o2, %o2
 csum_partial_fix_aligned:
 	brz,pt	%o3, 3f					! none to do
@@ -115,9 +117,9 @@ csum_partial_fix_aligned:
 	CSUM_BIGCHUNK(%o0, 0x20, %o2, %o4, %o5, %g2, %g3, %g4, %g5)
 	CSUM_BIGCHUNK(%o0, 0x40, %o2, %o4, %o5, %g2, %g3, %g4, %g5)
 	CSUM_BIGCHUNK(%o0, 0x60, %o2, %o4, %o5, %g2, %g3, %g4, %g5)
-	sub	%o3, 128, %o3				! detract from loop iters
 	addc	%g0, %o2, %o2				! sink in final carry
-	brnz,pt	%o3, 5b					! more to do
+	subcc	%o3, 128, %o3				! detract from loop iters
+	bne,pt	%icc, 5b				! more to do
 	 add	%o0, 128, %o0				! advance buf ptr
 3:	brz,pn	%g1, cpte				! nope
 	 andcc	%o1, 0xf, %o3				! anything left at all?
@@ -125,7 +127,7 @@ csum_partial_fix_aligned:
 	srl	%g1, 1, %o4				! compute offset
 	sub	%g7, %g1, %g7				! adjust jmp ptr
 	sub	%g7, %o4, %g7				! final jmp ptr adjust
-	jmp	%g7 + (cpte - 8 - 10b)			! enter the table
+	jmp	%g7 + (11f-10b)				! enter the table
 	 add	%o0, %g1, %o0				! advance buf ptr
 cptbl:	CSUM_LASTCHUNK(%o0, 0x68, %o2, %g2, %g3, %g4, %g5)
 	CSUM_LASTCHUNK(%o0, 0x58, %o2, %g2, %g3, %g4, %g5)
@@ -134,8 +136,8 @@ cptbl:	CSUM_LASTCHUNK(%o0, 0x68, %o2, %g2, %g3, %g4, %g5)
 	CSUM_LASTCHUNK(%o0, 0x28, %o2, %g2, %g3, %g4, %g5)
 	CSUM_LASTCHUNK(%o0, 0x18, %o2, %g2, %g3, %g4, %g5)
 	CSUM_LASTCHUNK(%o0, 0x08, %o2, %g2, %g3, %g4, %g5)
-	addc	%g0, %o2, %o2				! fetch final carry
-	andcc	%o1, 0xf, %g0				! anything left at all?
+11:	addc	%g0, %o2, %o2				! fetch final carry
+	andcc	%o1, 0xf, %o3				! anything left at all?
 cpte:	brnz,pn	%o3, csum_partial_end_cruft		! yep, handle it
 	 sethi	%uhi(KERNBASE), %g4
 	mov	%o2, %o0				! return computed csum
@@ -322,13 +324,14 @@ __csum_partial_copy_sparc_generic:
 	andcc	%o0, 0x4, %g0
 	or	%g3, %g7, %g7
 1:	be,pt	%icc, 3f
-	 andn	%g1, 0x7f, %g0
+	 andn	%g1, 0x7f, %g2
 	EX(ld	[%o0 + 0x00], %g4, add %g1, 0,#)
 	sub	%g1, 4, %g1
 	EX2(st	%g4, [%o1 + 0x00],#)
 	add	%o0, 4, %o0
 	addcc	%g4, %g7, %g7
 	add	%o1, 4, %o1
+	andn	%g1, 0x7f, %g2
 	addc	%g0, %g7, %g7
 cc_dword_aligned:
 3:	brz,pn	%g2, 3f			! nope, less than one loop remains
@@ -365,7 +368,7 @@ cctbl:	CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x68,%g2,%g3,%g4,%g5)
 	CSUMCOPY_LASTCHUNK(%o0,%o1,%g7,0x08,%g2,%g3,%g4,%g5)
 12:	EXT(cctbl, 12b, 22f,#)		! note for exception table handling
 	addc	%g0, %g7, %g7
-	andcc	%o3, 0xf, %g0		! check for low bits set
+	andcc	%g1, 0xf, %o3		! check for low bits set
 ccte:	bne,pn	%icc, cc_end_cruft	! something left, handle it out of band
 	 sethi	%uhi(KERNBASE), %g4	! restore gfp
 	mov	%g7, %o0		! give em the computed checksum
@@ -555,7 +558,7 @@ __csum_partial_copy_end:
 	add	%i1, %i2, %i1
 2:
 	mov	%i1, %o0
-        wr	%%g0, ASI_S, %%asi
+        wr	%g0, ASI_S, %asi
         call	__bzero_noasi
 	 mov	%i3, %o1
 1:
diff --git a/arch/sparc64/lib/copy_from_user.S b/arch/sparc64/lib/copy_from_user.S
index ba26a1c01..50ec7bb3d 100644
--- a/arch/sparc64/lib/copy_from_user.S
+++ b/arch/sparc64/lib/copy_from_user.S
@@ -15,12 +15,16 @@
 
 #include <asm/ptrace.h>
 #include <asm/asi.h>
+#include <asm/head.h>
+
+#define PRE_RETL sethi %uhi(KERNBASE), %g4; sllx %g4, 32, %g4;
 
 #define EX(x,y,a,b,z) 				\
 98: 	x,y;					\
 	.section .fixup,z##alloc,z##execinstr;	\
 	.align	4;				\
-99:	retl;					\
+99:	PRE_RETL				\
+	retl;					\
 	 a, b, %o0;				\
 	.section __ex_table,z##alloc;		\
 	.align	4;				\
@@ -33,6 +37,7 @@
 	.section .fixup,z##alloc,z##execinstr;	\
 	.align	4;				\
 99:	c, d, e;				\
+	PRE_RETL				\
 	retl;					\
 	 a, b, %o0;				\
 	.section __ex_table,z##alloc;		\
@@ -234,6 +239,7 @@ copy_user_last7:
 	EX(lduba [%o1] %asi, %g2, add %g0, 1,#)
 	stb	%g2, [%o0]
 1:
+	PRE_RETL
 	retl
  	 clr	%o0
 
@@ -332,6 +338,7 @@ short_table_end:
 	EX(lduba [%o1] %asi, %g2, add %g0, 1,#)
 	stb	%g2, [%o0]
 1:
+	PRE_RETL
 	retl
  	 clr	%o0
 
@@ -355,6 +362,7 @@ short_aligned_end:
 	.section .fixup,#alloc,#execinstr
 	.align	4
 97:
+	PRE_RETL
 	retl
 	 mov	%o2, %o0
 /* exception routine sets %g2 to (broken_insn - first_insn)>>2 */
@@ -388,6 +396,7 @@ short_aligned_end:
 1:
 	and	%g1, 0x7f, %o0
 	add	%o0, %g7, %o0
+	PRE_RETL
 	retl
 	 sub	%o0, %g2, %o0
 51:
@@ -413,6 +422,7 @@ short_aligned_end:
 3:
 	sll	%g2, 2, %g2
 2:
+	PRE_RETL
 	retl
 	 add	%g1, %g2, %o0
 52:
@@ -431,6 +441,7 @@ short_aligned_end:
 	add	%g2, %g4, %g2
 	and	%o2, 0xf, %o0
 	add	%o0, %o3, %o0
+	PRE_RETL
 	retl
 	 sub	%o0, %g2, %o0
 54:
@@ -441,6 +452,7 @@ short_aligned_end:
 	and	%o2, 0xf, %o2
 	sub	%o3, %o1, %o3
 	sub	%o2, %o4, %o2
+	PRE_RETL
 	retl
 	 add	%o2, %o3, %o0
 55:
@@ -452,5 +464,6 @@ short_aligned_end:
 	and	%g2, 1, %g2
 	sll	%o1, 1, %o1
 	add	%o2, %g2, %o0
+	PRE_RETL
 	retl
 	 add	%o0, %o1, %o0
diff --git a/arch/sparc64/lib/copy_to_user.S b/arch/sparc64/lib/copy_to_user.S
index 47a6bd337..733953743 100644
--- a/arch/sparc64/lib/copy_to_user.S
+++ b/arch/sparc64/lib/copy_to_user.S
@@ -14,13 +14,17 @@
  */
 
 #include <asm/ptrace.h>
+#include <asm/head.h>
 #include <asm/asi.h>
 
+#define PRE_RETL sethi %uhi(KERNBASE), %g4; sllx %g4, 32, %g4;
+
 #define EX(x,y,a,b,z) 				\
 98: 	x,y;					\
 	.section .fixup,z##alloc,z##execinstr;	\
 	.align	4;				\
-99:	retl;					\
+99:	PRE_RETL				\
+	retl;					\
 	 a, b, %o0;				\
 	.section __ex_table,z##alloc;		\
 	.align	4;				\
@@ -33,6 +37,7 @@
 	.section .fixup,z##alloc,z##execinstr;	\
 	.align	4;				\
 99:	c, d, e;				\
+	PRE_RETL				\
 	retl;					\
 	 a, b, %o0;				\
 	.section __ex_table,z##alloc;		\
@@ -234,6 +239,7 @@ copy_user_last7:
 	ldub	[%o1], %g2
 	EX(stba	%g2, [%o0] %asi, add %g0, 1,#)
 1:
+	PRE_RETL
 	retl
  	 clr	%o0
 
@@ -332,6 +338,7 @@ short_table_end:
 	ldub	[%o1], %g2
 	EX(stba	%g2, [%o0] %asi, add %g0, 1,#)
 1:
+	PRE_RETL
 	retl
  	 clr	%o0
 
@@ -355,6 +362,7 @@ short_aligned_end:
 	.section .fixup,#alloc,#execinstr
 	.align	4
 97:
+	PRE_RETL
 	retl
 	 mov	%o2, %o0
 /* exception routine sets %g2 to (broken_insn - first_insn)>>2 */
@@ -388,6 +396,7 @@ short_aligned_end:
 1:
 	and	%g1, 0x7f, %o0
 	add	%o0, %g7, %o0
+	PRE_RETL
 	retl
 	 sub	%o0, %g2, %o0
 51:
@@ -413,6 +422,7 @@ short_aligned_end:
 3:
 	sll	%g2, 2, %g2
 2:
+	PRE_RETL
 	retl
 	 add	%g1, %g2, %o0
 52:
@@ -431,6 +441,7 @@ short_aligned_end:
 	add	%g2, %g4, %g2
 	and	%o2, 0xf, %o0
 	add	%o0, %o3, %o0
+	PRE_RETL
 	retl
 	 sub	%o0, %g2, %o0
 54:
@@ -441,6 +452,7 @@ short_aligned_end:
 	and	%o2, 0xf, %o2
 	sub	%o3, %o1, %o3
 	sub	%o2, %o4, %o2
+	PRE_RETL
 	retl
 	 add	%o2, %o3, %o0
 55:
@@ -452,5 +464,6 @@ short_aligned_end:
 	and	%g2, 1, %g2
 	sll	%o1, 1, %o1
 	add	%o2, %g2, %o0
+	PRE_RETL
 	retl
 	 add	%o0, %o1, %o0
diff --git a/arch/sparc64/lib/strlen_user.S b/arch/sparc64/lib/strlen_user.S
index 24bea73fd..30beee3ff 100644
--- a/arch/sparc64/lib/strlen_user.S
+++ b/arch/sparc64/lib/strlen_user.S
@@ -8,6 +8,8 @@
  * Copyright (C) 1996,1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
  */
 
+#include <asm/asi.h>
+
 #define LO_MAGIC 0x01010101
 #define HI_MAGIC 0x80808080
 
@@ -19,21 +21,21 @@ __strlen_user:
 	be,pt	%icc, 9f
 	 sethi	%hi(HI_MAGIC), %o4
 10:
-	ldub	[%o0], %o5
+	lduba	[%o0] ASI_S, %o5
 	brz,pn	%o5, 21f
 	 add	%o0, 1, %o0
 	andcc	%o0, 3, %g0
 	be,pn	%icc, 4f
 	 or	%o4, %lo(HI_MAGIC), %o3
 11:
-	ldub	[%o0], %o5
+	lduba	[%o0] ASI_S, %o5
 	brz,pn	%o5, 22f
 	 add	%o0, 1, %o0
 	andcc	%o0, 3, %g0
 	be,pt	%icc, 5f
 	 sethi	%hi(LO_MAGIC), %o4
 12:
-	ldub	[%o0], %o5
+	lduba	[%o0] ASI_S, %o5
 	brz,pn	%o5, 23f
 	 add	%o0, 1, %o0
 	ba,pt	%icc, 13f
@@ -45,7 +47,7 @@ __strlen_user:
 5:
 	or	%o4, %lo(LO_MAGIC), %o2
 13:
-	ld	[%o0], %o5
+	lda	[%o0] ASI_S, %o5
 2:
 	sub	%o5, %o2, %o4
 	andcc	%o4, %o3, %g0
@@ -68,7 +70,7 @@ __strlen_user:
 	andcc	%o5, 0xff, %g0
 	bne,a,pt %icc, 2b
 14:
-	 ld	[%o0], %o5
+	 lda	[%o0] ASI_S, %o5
 	add	%o4, 1, %o4
 1:
 	retl
diff --git a/arch/sparc64/lib/strncpy_from_user.S b/arch/sparc64/lib/strncpy_from_user.S
index 05a48eb5a..e0fb0f09b 100644
--- a/arch/sparc64/lib/strncpy_from_user.S
+++ b/arch/sparc64/lib/strncpy_from_user.S
@@ -24,14 +24,14 @@ __strncpy_from_user:
 	sub	%g0, %o2, %o3
 	add	%o0, %o2, %o0
 10:
-	ldub	[%o1 + %o3], %o4
+	lduba	[%o1 + %o3] ASI_S, %o4
 1:
 	brz,pn	%o4, 2f
 	 stb	%o4, [%o0 + %o3]
 	addcc	%o3, 1, %o3
 	bne,pt	%xcc, 1b
 11:
-	 ldub	[%o1 + %o3], %o4
+	 lduba	[%o1 + %o3] ASI_S, %o4
 	retl
 	 mov	%o2, %o0
 2:
diff --git a/arch/sparc64/mm/asyncd.c b/arch/sparc64/mm/asyncd.c
index 4e7de16fb..0272b09c2 100644
--- a/arch/sparc64/mm/asyncd.c
+++ b/arch/sparc64/mm/asyncd.c
@@ -1,4 +1,4 @@
-/*  $Id: asyncd.c,v 1.1 1996/12/26 10:24:24 davem Exp $
+/*  $Id: asyncd.c,v 1.2 1997/05/15 21:14:32 davem Exp $
  *  The asyncd kernel daemon. This handles paging on behalf of 
  *  processes that receive page faults due to remote (async) memory
  *  accesses. 
@@ -153,7 +153,7 @@ static int fault_in_page(int taskid,
 	if(!pte)
 		goto no_memory;
 	if(!pte_present(*pte)) {
-		do_no_page(tsk, vma, address, write);
+		handle_mm_fault(tsk, vma, address, write);
 		goto finish_up;
 	}
 	set_pte(pte, pte_mkyoung(*pte));
@@ -165,12 +165,11 @@ static int fault_in_page(int taskid,
 		flush_tlb_page(vma, address);
 		goto finish_up;
 	}
-	do_wp_page(tsk, vma, address, write);
+	handle_mm_fault(tsk, vma, address, write);
 
 	/* Fall through for do_wp_page */
 finish_up:
 	stats.success++;
-	update_mmu_cache(vma, address, *pte);
 	return 0;
 
 no_memory:
diff --git a/arch/sparc64/mm/fault.c b/arch/sparc64/mm/fault.c
index 0dd118c8e..dc28ac339 100644
--- a/arch/sparc64/mm/fault.c
+++ b/arch/sparc64/mm/fault.c
@@ -1,4 +1,4 @@
-/* $Id: fault.c,v 1.4 1997/03/11 17:37:07 jj Exp $
+/* $Id: fault.c,v 1.8 1997/05/18 04:16:52 davem Exp $
  * arch/sparc64/mm/fault.c: Page fault handlers for the 64-bit Sparc.
  *
  * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
@@ -134,8 +134,11 @@ asmlinkage int lookup_fault(unsigned long pc, unsigned long ret_pc,
 	return 0;
 }
 
+/* #define FAULT_TRACER */
+
 asmlinkage void do_sparc64_fault(struct pt_regs *regs, int text_fault, int write,
-				 unsigned long address)
+				 unsigned long address, unsigned long tag,
+				 unsigned long sfsr)
 {
 	struct vm_area_struct *vma;
 	struct task_struct *tsk = current;
@@ -143,7 +146,19 @@ asmlinkage void do_sparc64_fault(struct pt_regs *regs, int text_fault, int write
 	unsigned long fixup;
 	unsigned long g2;
 	int from_user = !(regs->tstate & TSTATE_PRIV);
-
+#ifdef FAULT_TRACER
+	static unsigned long last_addr = 0;
+	static int rcnt = 0;
+
+	printk("do_sparc64_fault(PC[%016lx],t[%d],w[%d],addr[%016lx]tag[%016lx]"
+	       "sfar[%016lx])\n", regs->tpc, text_fault, write, address, tag, sfsr);
+	if(address == last_addr && rcnt++ > 5) {
+		printk("Wheee lotsa bogus faults, something wrong, spinning\n");
+		while(1)
+			barrier();
+	}
+	last_addr = address;
+#endif
 	lock_kernel ();
 	down(&mm->mmap_sem);
 	vma = find_vma(mm, address);
@@ -168,7 +183,7 @@ good_area:
 		if(!(vma->vm_flags & (VM_READ | VM_EXEC)))
 			goto bad_area;
 	}
-	handle_mm_fault(vma, address, write);
+	handle_mm_fault(current, vma, address, write);
 	up(&mm->mmap_sem);
 	goto out;
 	/*
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 57ca5eb92..cf378a266 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -1,4 +1,4 @@
-/*  $Id: init.c,v 1.24 1997/04/17 21:49:41 jj Exp $
+/*  $Id: init.c,v 1.28 1997/05/18 04:16:53 davem Exp $
  *  arch/sparc64/mm/init.c
  *
  *  Copyright (C) 1996,1997 David S. Miller (davem@caip.rutgers.edu)
@@ -457,16 +457,18 @@ void sparc_ultra_unmapioaddr(unsigned long virt_addr)
 	pte_clear(ptep);
 }
 
-#ifdef DEBUG_MMU
 void sparc_ultra_dump_itlb(void)
 {
         int slot;
 
-        prom_printf ("Contents of itlb:\n");
-        for (slot = 0; slot < 64; slot+=2) {
-        	prom_printf ("%2x:%016lx,%016lx    %2x:%016lx,%016lx\n", 
+        printk ("Contents of itlb: ");
+	for (slot = 0; slot < 14; slot++) printk ("    ");
+	printk ("%2x:%016lx,%016lx\n", 0, spitfire_get_itlb_tag(0), spitfire_get_itlb_data(0));
+        for (slot = 1; slot < 64; slot+=3) {
+        	printk ("%2x:%016lx,%016lx %2x:%016lx,%016lx %2x:%016lx,%016lx\n", 
         		slot, spitfire_get_itlb_tag(slot), spitfire_get_itlb_data(slot),
-        		slot+1, spitfire_get_itlb_tag(slot+1), spitfire_get_itlb_data(slot+1));
+        		slot+1, spitfire_get_itlb_tag(slot+1), spitfire_get_itlb_data(slot+1),
+        		slot+2, spitfire_get_itlb_tag(slot+2), spitfire_get_itlb_data(slot+2));
         }
 }
 
@@ -474,14 +476,16 @@ void sparc_ultra_dump_dtlb(void)
 {
         int slot;
 
-        prom_printf ("Contents of dtlb:\n");
-        for (slot = 0; slot < 64; slot+=2) {
-        	prom_printf ("%2x:%016lx,%016lx    %2x:%016lx,%016lx\n", 
+        printk ("Contents of dtlb: ");
+	for (slot = 0; slot < 14; slot++) printk ("    ");
+	printk ("%2x:%016lx,%016lx\n", 0, spitfire_get_dtlb_tag(0), spitfire_get_dtlb_data(0));
+        for (slot = 1; slot < 64; slot+=3) {
+        	printk ("%2x:%016lx,%016lx %2x:%016lx,%016lx %2x:%016lx,%016lx\n", 
         		slot, spitfire_get_dtlb_tag(slot), spitfire_get_dtlb_data(slot),
-        		slot+1, spitfire_get_dtlb_tag(slot+1), spitfire_get_dtlb_data(slot+1));
+        		slot+1, spitfire_get_dtlb_tag(slot+1), spitfire_get_dtlb_data(slot+1),
+        		slot+2, spitfire_get_dtlb_tag(slot+2), spitfire_get_dtlb_data(slot+2));
         }
 }
-#endif
 
 /* paging_init() sets up the page tables */
 
@@ -643,7 +647,7 @@ __initfunc(void mem_init(unsigned long start_mem, unsigned long end_mem))
 	high_memory = (void *) end_mem;
 
 	start_mem = PAGE_ALIGN(start_mem);
-	num_physpages = (start_mem - phys_base - PAGE_OFFSET) >> PAGE_SHIFT;
+	num_physpages = (start_mem - PAGE_OFFSET) >> PAGE_SHIFT;
 
 	addr = PAGE_OFFSET;
 	while(addr < start_mem) {
@@ -694,6 +698,11 @@ __initfunc(void mem_init(unsigned long start_mem, unsigned long end_mem))
 		min_free_pages = 16;
 	free_pages_low = min_free_pages + (min_free_pages >> 1);
 	free_pages_high = min_free_pages + min_free_pages;
+
+#if 0
+	printk("Testing fault handling...\n");
+	*(char *)0x00000deadbef0000UL = 0;
+#endif
 }
 
 void free_initmem (void)
@@ -702,9 +711,14 @@ void free_initmem (void)
 	
 	addr = (unsigned long)(&__init_begin);
 	for (; addr < (unsigned long)(&__init_end); addr += PAGE_SIZE) {
-		mem_map[MAP_NR(addr)].flags &= ~(1 << PG_reserved);
-		atomic_set(&mem_map[MAP_NR(addr)].count, 1);
-		free_page(addr);
+		unsigned long page = addr;
+
+		if(page < ((unsigned long)__va(phys_base)))
+			page += phys_base;
+
+		mem_map[MAP_NR(page)].flags &= ~(1 << PG_reserved);
+		atomic_set(&mem_map[MAP_NR(page)].count, 1);
+		free_page(page);
 	}
 }